Report DEQ08-LAB-0048-TR

PREDATOR: Development and use of RIVPACS-type macroinvertebrate models to assess the biotic condition of wadeable Oregon streams (November 2005 models)

By: Shannon Hubler

July 2008

Last Update 07/14/2008 DEQ08-LAB-0048-TR Version 1.1

Web pub#: 10-LAB-004

This report prepared by: Oregon Department of Environmental Quality Laboratory and Environmental Assessment Division Watershed Assessment Section 3150 NW 229th, Suite 150, Hillsboro, Oregon 97124 U.S.A. 1-800-452-4011 www.oregon.gov/deq

Contact: Shannon Hubler (503) 693-5728

1

List of Tables ...... 3 List of Figures ...... 3 Rationale ...... 4 What is a Predictive Model? ...... 4 Why Macroinvertebrates? ...... 4 The PREDictive Assessment Tool for Oregon (PREDATOR) ...... 5 How does a predictive model differ from a Multi-metric approach? ...... 5 Developing the Models ...... 5 Macroinvertebrate Sampling Protocols ...... 6 Taxonomy ...... 6 Model Development ...... 8 Null Models ...... 9 Final Model Selection ...... 10 Assessing model quality ...... 13 Comparisons to other PNW RIVPACS-type models...... 13 Null model performance ...... 16 Using the models...... 16 PREDATOR outputs ...... 16 Benchmarks of biological condition ...... 19 Population Assessments ...... 20 Individual site assessments ...... 21 Causes of poor biological condition ...... 23 The importance of assessing multiple assemblages ...... 25 Conclusions ...... 26 Future versions of PREDATOR ...... 26 Recommendations and Needs ...... 27 Acknowledgements ...... 28 Literature Cited ...... 28 Appendix A...... 31 Appendix B...... 44 Appendix C...... 46

2

List of Tables Table 1. A hypothetical example of how consistent taxonomic levels are achieved...... 7 Table 2. PREDATOR model specifications for three regions in Oregon...... 11 Table 3. PREDATOR 2005 model performance statistics...... 14 Table 4. O/E benchmarks for describing biological condition for predictive PREDATOR models...... 19 Table 5. Benchmarks for describing biological condition for the null PREDATOR model...... 19 Table 6. OTUs and phylogenetic classifications used in PREDATOR models...... 31 Table 7. Candidate predictor variables that were examined in PREDATOR model development...... 44 Table 8. MWCF reference sites and corresponding environmental data ...... 46 Table 9. WC+CP reference sites and corresponding environmental data ...... 47

List of Figures Figure 1. PREDATOR consists of two predictive models and one null model ...... 12 Figure 2. Performance of the MWCF model and of the WC+CP model...... 15 Figure 3. Frequency distributions of O/E scores for samples assessed by the MWCF model...... 18 Figure 4. The extent of biotic condition classes for samples in the Coast Range ecoregion and the Willamette Valley ecoregion ...... 22 Figure 5. Identifying potential causes of impairment in two sites with O/E in most disturbed condition...... 24

3

Rationale The Oregon Department of Environmental Quality (DEQ) is responsible for protecting the waters of the state from pollution that may adversely affect drinking water, aquatic life and recreational uses. DEQ routinely monitors conventional water quality parameters such as nutrients, dissolved oxygen, pH, turbidity, conductivity and bacteria to report on the water quality status and trends in Oregon. However, resource limitations make it impractical to measure all the potential pollutants which may impair Oregon’s waters. Aquatic insect communities are direct indicators of biological conditions and a surrogate for watershed health. They provide a cost effective screening tool for assessing and identifying problems that may require further examination.

The purpose of this document is to provide a background on predictive modelling, its utility, and the specific application of the macroinvertebrate models used by the Oregon DEQ. What is a Predictive Model? A predictive model, in this case, is a tool used to assess the integrity of an aquatic insect assemblage. Predictive modelling estimates the expected occurrence of macroinvertebrates at a sample location. This is done by developing a list of insect species that commonly occur at least disturbed, or reference, locations that have similar natural characteristic to the sample locations. The list of species generated from the reference locations is known as the “Expected” taxa list or “E”. This list is compared to the captured aquatic insects or,“Observed” taxa (“O”), at an assessment site. The predictive model output is the observed to expected (O/E) taxa ratio. Scores less than one have fewer taxa at a site than were predicted by the model. Scores greater than one are either equivalent to the reference location or may have an enhanced insect community as a result of some type of enrichment.

Another way to think of the score is in terms of the percentage of taxa loss or gain. Values less than 1.0 represent a loss of common native reference taxa. Percent taxa loss or gain is defined as: (O/E – 1.0) * 100

A negative value means a sample has lost reference taxa, while a positive value means the sample has gained reference taxa

Why Macroinvertebrates? Macroinvertebrates include freshwater insects, crustaceans, mollusks, bivalves and other invertebrates larger than one half millimeter in size. They are important because they occupy a central role in food chains and ecosystem processes (Wallace and Webster 1996). Macroinvertebrates are easy to collect, are relatively cheap to process and analyze, and show strong responses to many stressors. These benefits, make macroinvertebrates the most commonly used aquatic organisms for assessing stream biological integrity. For a thorough examination of the role of macroinvertebrates in assessing biological integrity, see Rosenberg and Resh (1993) and Wright et. al (2000).

4

The PREDictive Assessment Tool for Oregon (PREDATOR) PREDATOR consists of three regional models that assess the biological integrity of wadeable streams across Oregon. DEQ developed the models to supply a scientifically rigorous bioassessment tool that is easy to apply and provides a more complete understanding of the stream conditions across Oregon.

Similar to other predictive models, PREDATOR, generates an expected occurrence probability (how likely a taxon is to occur) for each species at a test site. Common taxa at reference sites with similar environmental conditions will have higher occurrence probabilities at test sites. The sum of the occurrence probabilities is the expected number of reference taxa, “E”. Expected taxa are restricted to those taxa that were found at reference sites used for building the model. In PREDATOR, only taxa with occurrence probabilities greater than 50% were used to calculate the expected taxa list (see Ostermiller and Hawkins 2004). The observed taxa, or “O”, are the number of expected taxa that were actually collected at the test site.

The first predictive model of this kind, the River InVertebrate Prediction and Classification System (RIVPACS), was developed in Britain in the 1980’s. Two excellent overviews of the RIVPACS approach to predictive modeling are the Western Center for Monitoring (2006) and the Centre for Ecology and Hydrology (2006). For a more detailed discussion of predictive models (and other bioassessment techniques), see Wright et al (2000).

How does a predictive model differ from a Multi-metric approach? Another common method for assessing biological conditions is called the multi-metric approach. Unlike the predictive modeling approach which uses raw data, the multi- metric approach summarizes biological data into groups. These groups are based on knowledge about the life history of the assessment organisms and convey information about the conditions of the river or stream they live in. Biological metrics are often selected to provide information on different aspects of the stream conditions or species composition like temperature preferences, nutrient preferences, the percentage of alien species, functional feeding groups, etc. The metrics are analyzed to determine which ones are most sensitive to disturbance. Once the metrics are selected, they are added together to create single score called an Index of Biological Integrity, or IBI. The final score is an index value for making comparisons with reference sites and other locations. Unlike the predictive modeling approach it does not represent the loss or gain of taxa at a site.

DEQ uses a multi-metric method for assessing fish and aquatic vertebrate assemblages. In the past, DEQ used the multi-metric approach to assess macroinvertebrate assemblages in Oregon but these indexes were developed using smaller-scale datasets which limited their applicability to other areas of the state.

Developing the Models There were five main steps to developing the PREDATOR models: 1) Establishing consistent macroinvertebrate sampling protocols and collection periods

5

2) Selecting regional reference sites 3) Grouping reference sites based on the macroinvertebrate communities 4) Relating reference groups to predictor variables 5) Assessing model performance

Macroinvertebrate Sampling Protocols Macroinvertebrate data from three sources, Utah State University (USU), the Washington Department of Ecology (WDOE) and DEQ were used to develop PREDATOR. In order to use data from multiple sources, consistent sampling protocols were necessary. A consistent level of sampling effort is important for developing precise predictive models. While there were slight differences in sampling methods, all three sources collected with the same sampling effort. DEQ macroinvertebrate sampling followed the standard methods described in the DEQ Mode of Operations Manual (DEQ 2004).

Basic sampling methods included the following:

Sampling season: Summer low flow (late June to early October) Habitat: fast water (riffle) Sample location: systematic random selection Collection device: D-frame kicknet, 500 μm mesh Sample area: DEQ: 1998-2002 = 2 ft x 1 ft (4 composited kicks); 2003-2004 = 1 ft x 1 ft (8 composited kicks) USU: 1998-2004 = 1 ft x 1 ft (8 composited kicks) WDOE: 1998-2004 = 2 ft x 1 ft (4 composited kicks)

Total area: 8-ft2, one composite sample Laboratory Sub-sample: max. 500 individuals; 10x magnification Identification: typically genus/species; Chironomidae to sub- family/tribe

Taxonomy Predicative models require a consistent level of taxonomy be applied to all samples used to build and assess the models (Moss et. al 1999, Ostermiller and Hawkins 2004). Some more highly resolved taxa were aggregated to a less resolved taxonomic category (e.g., all species in one genus were grouped together). Alternatively, some less resolved taxa were excluded from analyses and more highly resolved taxa were retained (e.g., specimens that were only able to be identified to family were deleted from datasets if the vast majority of individuals within that same family were able to be identified to the genus level). A hypothetical example of this procedure is shown below (Table 1). These exercises resulted in a list of operational taxonomic units (OTUs) that vary in their level of taxonomic resolution, but are unique from one another (no ambiguous taxa).

In the example presented in Table 1, the less resolved “Baetidae” were dropped from the analyses because there were few individuals identified to this level, plus there were many individuals identified to more highly resolved (genus and species) levels. In contrast, all the species level identifications under the genera “Baetis” were aggregated up to the

6

genus level. Again, in this case more information (individuals) was available at the genus level than at the species level. Also, note that “Diphetor hageni” is at the species level, while the other two taxa (“Baetis” and “Acentrella”) were kept at the genus level. This is acceptable, as long as each taxon is unique from all other taxa.

Table 1. A hypothetical example of how consistent taxonomic levels are achieved. (“Lowest level identification” = the lowest taxonomic level achieved by an expert taxonomist. “Sample abundance” = the number of individuals collected at a site. “Unique taxa” = a taxonomic level where there are no individuals in a sample at a lower, related taxonomic level. )

Site Lowest level Taxonomic Sample Unique Action Operational Model name identification level Abundance Taxa Taxonomic abundance Units Fox Baetidae Family 7 No Exclude -- -- Creek Fox Baetis Genus 132 No None Baetis 150 Creek Fox Baetis Species 13 Yes Aggregate Baetis -- Creek tricaudatus to genus Fox Baetis Species 2 Yes Aggregate Baetis -- Creek bicaudatus to genus Fox Baetis alius Species 3 Yes Aggregate Baetis -- Creek to genus Fox Acentrella Genus 15 Yes None Acentrella 15 Creek Fox Diphetor Species 4 Yes None Diphetor 4 Creek hageni hageni

A table of OTUs and phylogenetic classification for the November 2005 PREDATOR models is shown in Appendix A. The full table is available for download from the Western Center for Monitoring (2006). Future versions of PREDATOR are likely to have differing levels of taxonomy for certain groups (e.g., chironomidae) than the current models. Our objectives will always be to increase taxonomic information as much as possible, while maintaining high model performance. Each subsequent version of PREDATOR will include full documentation of OTU levels.

The goal of assigning OTUs is to retain as much taxonomic information as possible. Optimally, all taxa in a sample would be identified to species, accounting for differing ecological requirements. However, the taxonomic literature allows for only certain groups of taxa to be identified to species. Additionally, some laboratories routinely identify certain groups of taxa to less resolved levels than others, limiting the taxonomic level to which those groups can be pooled across labs. For instance, in the PREDATOR models we combined data from three separate taxonomic laboratories. The trichoptera genus Rhyacophila was identified to species group for two of these laboratories, but to the less resolved genus for the third laboratory (47% of reference samples, and most of the eastern Oregon reference samples). There do appear to be differing ecological requirements among the species groups which could be useful in the models, but using the species group as the OTU for this set of taxa would have resulted in throwing out all individuals identified to the genus level. This would mean that all information related to

7

Rhyacophila would be discarded for ~47% of the reference sites used in development of RIVPACS models. The real question then becomes: Is the information at the genus level of Rhyacophila more important to retain for all reference sites? Or are the species groups so different in their ecological requirements that it makes more sense to have no information for this group at 47% of reference sites. Local experts and DEQ databases of ecological traits were consulted when ultimately making decisions on whether to aggregate or discard taxa from the models.

Next, the 500 organism count from each sample is randomly sub-sampled to 300 individuals using a simple computer routine. Model precision and accuracy was shown to increase with sub-sampled counts up to ~300 individuals (Ostermiller and Hawkins 2004). The purpose of this sub-sampling routine is to standardize the effort across samples. Species richness metrics such as O/E are highly correlated to the total amount of sample sorted, thus samples with more individuals have a greater likelihood of having higher O/E scores. By standardizing the sub-sample count to 300, we are attempting to even out the effects of differing sample sizes. Reference samples that contained less than 200 OTU individuals were excluded from the model building process. Samples may not have had at least 300 individuals due to either low productivity (naturally low in macroinvertebrate abundance) or many individuals were dropped because they were not identified to the appropriate taxonomic level (OTU). The “subsample.exe” program is available for download at the WCM website (Western Center for Monitoring 2006).

Macroinvertebrate assemblages were sampled during the summer months (June through early October) from 1998-2004. Sample sites included a wide range of wadeable stream types and span nearly all of the major ecoregions in the State of Oregon. Sites were surveyed either as part of random probabilistic surveys or as hand-picked reference sites using best professional judgment. During field collection, sites were also screened for approximately 30 human activities at the reach scale. Those activities closer to the stream bank were assigned higher scores. Post-sampling, all sites were screened based on the degree of human activities in their drainage areas for road density, urban and agricultural use, and active or recent logging (Drake 2004). A total of 205 reference sites were chosen for model calibration; 125 of the sites were sampled by DEQ, 96 were sampled by Utah State University (USU), and 6 were sampled by Washington Department of Ecology (WDOE). We included sites from Washington State to allow for assessment of conditions in the Columbia Plateau ecoregion, where DEQ has not currently identified any reference sites. For a thorough discussion of the reference condition approach, see Reynoldson et. al 1997 and Stoddard et. al 2006.

Model Development Reference sites were grouped according to the similarities of their sampled invertebrate assemblages. This was accomplished through cluster analysis, and is based entirely on the biology. Clustering was performed using the Sorenson dissimilarity distance measure and flexible beta linkage (β = -0.6) (McCune and Grace 2002, Van Sickle et. al 2006). The choice of beta level was made to reduce the amount of chaining in the resulting dendrogram and to aid in the identification of reference groups. Reference groups were identified by “pruning” the resulting dendrogram at a level that maximized with-in group fidelity, as well as group size (≥ 5 sites). All groups were formed by pruning across the dendrogram at the same height.

8

The resulting reference groups were then evaluated against numerous environmental variables to determine what non-anthropogenic factors best predicted reference group membership. These environmental variables and associated reference site groups create the basis for predictive models. With a predictive model a test site is assigned a likelihood of belonging to each reference group, based on the values of environmental (predictor) variables. The set of predictor variables that best explained differences in reference groups was determined through discriminant function analysis (DFA) (McCune and Grace 2002). All possible combinations of predictor variables (best-subsets) were screened (Van Sickle et al. 2006). Other statistical methods for predicting group membership, such as Classification and Regression Tree or Random Forests, were not explored. These methods offer promise in improving our attempts to model the environmental drivers of biological reference groups, but at the time we developed these models the literature in relation to stream bioassessment was sparse (Cao et. al 2007).

Only variables that are unlikely to be affected by human disturbance were used to determine the probability of a test site belonging to each reference group. Human disturbances can alter certain habitat or chemical variables, which can result in inaccurate predictions. (For example: Say a model uses conductivity as the only predictor variable and some reference groups have naturally low conductivity, while other reference groups have naturally high conductivity. If a test site we wish to assess with the model is in a stream with naturally low conductivity, but is immediately downstream of an irrigation return flow—which in this case artificially raises the conductivity—the model would inappropriately predict bugs at the test site similar to those bugs found at high conductivity reference sites.)

We examined a variety of predictor variables to see which ones best predicted the biological groups of the reference sites (Appendix B). We limited the variables to those that could be obtained from GIS coverages to minimize the amount of time and effort in the field. All that is required from a field crew is a macroinvertebrate sample and an accurate latitude and longitude. All other predictor information can be obtained in an office setting.

For a more detailed description of predictive modeling, as well as many literature resources, see the WCM website (Western Center for Monitoring 2006).

Null Models In some cases, it may not be advantageous to develop a predictive model. This can occur for a variety of reasons, such as too few reference sites or environmental variables that do not allow for more accurate predictions. A null model is an alternate approach that is not based on a prediction of reference taxa using multivariate statistics. A null model does not use any clustering of reference sites into groups. The expected taxa list (E) is the common reference taxa—those taxa that occur at greater than 50% of the reference sites used in the null model. “O” then is the taxa that were expected and collected at a site (the taxa that count towards O are limited to those that were included in E). Besides offering an assessment option when predictive modeling does not work, null models also provide a comparison to see how much precision and accuracy we gain through predictive modeling process. If our predictive models do not show a significant level of

9 improvement in model precision and accuracy, it is hard to justify the additional work required to make and use the predictive models.

Scoring a test sample using the null model is simple. The expected number of reference taxa (E) is always the same, because there is no predictive function. “E” is simply the sum of the frequency of occurrences of the ten taxa collected in 50% or more of the reference sites (Table 2). For the PREDATOR—NBR null model, E is always equal to 7.6. “O” then is the sum of how many of the ten reference taxa were observed in the sample.

Final Model Selection The scale at which models are developed and applied affects their accuracy and precision. Based on the number and location of reference sites in Oregon, we examined several approaches: 1) a single statewide model, 2) separate Eastern and Western Oregon models, and 3) Level II ecoregions. A single Oregon model covering the entire state lacked adequate precision. The standard deviation (SD) of reference site O/E scores in our statewide model was never less than 0.20. A SD ~ 0.17 is generally viewed as an acceptable target (Western Center for Monitoring 2006). Next we attempted to split the state into two regions, “East” and “West”, with the Cascades crest as the division. Model results were unsatisfactory compared with previous modeling attempts within Oregon, so we proceeded to examining smaller regional models. It was not possible to create level III ecoregion models due to the small sample sizes of reference sites in all but a few ecoregions, nor was it possible to create models based on river basins. Our best regional models approximated level II ecoregions. Models at this scale maximized both sample size and inclusion of as much of the state as possible. In the end, two predictive models and one null model were developed for Oregon (Table 2, Figure 1). (See Appendix C for a list of reference sites and environmental data used in each model.) The Marine Western Coastal Forest (MWCF) predictive model covers streams in the Coast Range and Willamette Valley ecoregions. The Western Cordillera + Columbia Plateau (WC+CP) predictive model covers streams in the Klamath Mountains, Cascades, East Cascades, Blue Mountains, and Columbia Plateau ecoregions. The Northern Basin and Range (NBR) null model covers streams in southeastern Oregon.

There are a few things to note about these models. First, the southeastern part of Oregon (Northern Basin and Range level III ecoregion) is assessed using a null model. We found that including these nine reference sites in any of our other predictive models significantly reduced model performance. Including samples from this region in any of our other models always resulted in reduced model performance. Second, the types of streams used to build the models were wadeable (typically first- through fourth-order) streams that contained fast water habitats (riffles). Third, the Columbia Plateau is not actually a part of the Western Cordillera level II ecoregion. It is a part of the same level II ecoregion (Western Interior Basin and Ranges) as the southeastern Oregon sites used in the null model. However, we found that including the Columbia Plateau reference sites with the reference sites from the Western Cordillera level II ecoregion did not reduce predictive model performance.

These three models cover all level III ecoregions in Oregon, except for the Snake River Plains in far eastern Oregon (Figure 1). Currently, DEQ does not have any reference

10 sites in this ecoregion. In the future, we plan to utilize reference sites in this ecoregion identified by the Idaho Department of Environmental Quality, as we did in the Columbia Plateau by using reference sites from Washington.

Table 2. PREDATOR model specifications for three regions in Oregon. (See Appendix A for variable descriptions.) Marine Western Western Cordillera + Northern Basin and Coastal Forest Columbia Plateau (WC+CP) Range (MWCF) (NBR) Type of Model Predictive Predictive Null Level 2 ecoregion + : Level 2 ecoregion: Level 3 ecoregion: Cascades, Klamath Mountains, Regions Coast Range & Northern Basin and East Cascades, Blue Willamette Valley Range Mountains, + Columbia Plateau Wadeable, Wadeable, Wadeable, Stream type fast water fast water fast water Eastern Oregon, elevation, Predictor Julian date, mean annual precipitation, None variables longitude annual maximum air temperature Reference 3 5 1 groups Temporal range 1998-2004 1998-2004 1999-2004 Occurrence Probability 0.5 0.5 0.5 (probability of capture Organism sample 300 300 300 count # Minimum # of 200 200 200 organisms # of reference 38 167 9 sites Baetis, Chironominae, Optioservus, Orthocladiinae, Null model taxa n/a n/a Rhyacophila, Trombidiformes, Diphetor_hageni, Epeorus, Zaitzevia, Brachycentrus

11

Figure 1. PREDATOR consists of two predictive models (1-Marine West Coast Forest, 2-Western Cordillera and Columbia Plateau) and one null model (Western Interior Basin and Range). No model exists for the Snake River Plains ecoregion.

12

Assessing model quality This section is provided for those interested in how DEQ assessed model accuracy and precision. More detailed information about measuring model performance can be found at the WCM website (Western Center for Monitoring 2006), Ostermiller and Hawkins (2004), and Van Sickle et. al (2005).

Statistical results of the three final models are shown in Table 3. The MWCF and WC+CP models showed substantial improvement over the null models for each region. The “Null model” represents the upper end of variability that our models try to improve upon. If a predictive model does not show significant reductions in model errors over a null model, then it does not make sense to make a more complicated model. “Replicate sampling error” represents the lowest amount of variability we can expect to achieve in our models. A good model will show results closer to the replicate sampling error than to the null model error. For a more thorough examination of the use of null model errors and replicate sampling errors as upper and lower baselines for model performance, see Van Sickle et al. (2005).

Accuracy and precision can be examined in several ways (Western Center for Monitoring 2006). One way to estimate model accuracy is to look at the mean O/E scores of the sites used to build the models. Accurate models have a mean O/E close to 1.0. All three of our models’ mean O/E values (for the reference sites used to build the models) are essentially 1.0 (Table3). Another way to examine model accuracy is to examine a plot of “O” versus “E” for the reference sites used to build the model. An accurate model should have a scatterplot that resembles a 1:1 line (i.e., a regression line slope close to 1.0 and an intercept close to 1.0). The slope of the regression lines for the MWCF model was 1.2 and the slope of the WC+CP model was 1.1 (Figure 2). Both models approximate the 1:1 line well.

Model precision can be estimated in two ways. One is to examine the spread of O/E scores in reference sites, represented by the standard deviation of O/E values. Precise models typically result in predictive model standard deviations of approximately 0.15 (Western Center for Monitoring 2006). The MWCF model was very precise with a predictive model standard deviation of 0.12, while the WC+CP model showed good precision with a model standard deviation of 0.15 (Table 3). Another way to examine precision is to look at the amount of variation in “O” that is predicted by “E”, which is represented by the r2 value from a regression of “O” to “E” at reference sites (Figure 2). In general, good models have r2 values between 0.5-0.75. The MWCF r2 (0.66) showed good precision, while the WC+CP r2 (0.33) suggests lower precision for this model.

Comparisons to other PNW RIVPACS-type models The predictive models developed for Oregon compare favorably to other RIVPACS-type models developed in the Pacific Northwest (PNW). Precision (measured as SD) of the WC+CP predictive model was similar to models created from Wyoming (Hargett et. al 2007) and all of Oregon (Van Sickle et. al 2006). The precision of the MWCF model was similar to a model for Western Oregon and Washington (Ostermiller and Hawkins 2004). For all of the PNW models, candidate predictor variables were fairly similar.

13

Sampling date, spatial location, stream size and power, geology, ecoregions, and climate variables were common to all modeling exercises. In all of these models, the trend seems to be to try to utilize predictor variables than can be derived through geographical information systems (GIS) exercises, rather than collecting more intensive field data. The number of final predictor variables used in the MWCF and WC+CP were lower than in Hargett et. al (2007) and Ostermiller and Hawkins (2004). In our models, we followed the results of Van Sickle et. al (2006) and used lower order (number of final variables) models in an attempt to avoid overfitting.

Table 3. PREDATOR 2005 model performance statistics. Model performance is shown for reference sites in level III ecoregions used to build the respective predictive models. “n” = sample size, “O/E” = observed/expected, and “SD” = standard deviation of O/E scores.

Model N Mean O/E SD Marine Western Coastal Forest 38

Null model 1.00 0.14

Predictive model 0.99 0.12 Replicate sampling error 0.11

Coast Range 28 0.98 0.12 Willamette Valley 10 1.04 0.14

Western Cordillera + Columbia Plateau 167

Null model 1.00 0.18 Predictive model 1.01 0.15

Replicate sampling error 0.13 Cascades 101 1.01 0.17

East Cascades 11 0.97 0.17

Klamath Mountains 10 0.99 0.13

Blue Mountains 39 1.02 0.12

Columbia Plateau 6 1.12 0.10 Northern Basin and Range 9

Null model 1.00 0.29

14

30

A 25

20 y = 1.2343x - 4.2737

Observed (O) Observed 2 R = 0.6645

15

10 10 15 20 25 30 25 Expected (E)

B 20

y = 1.0882x - 1.0884 2 15 R = 0.3328

Observed (O) Observed

10

5 5 10152025 Expected (E) Figure 2. Performance of the MWCF model (panel A) and of the WC+CP model (panel B). The solid line is the 1:1 line representing a perfect relationship between Observed (O) and Expected (E). The dashed line is the regression line between O and E. The r2 represents the percent of O explained by E.

15

In theory, the predictor variables used in development of RIVPACS models should be unaffected by anthropogenic disturbances. However, in practice, this is not always followed. Harget et. al (2007) and Ostermiller and Hawkins (2004) used substrate composition metrics as predictor variables, and they were also used in the original RIVPACS models developed in Britain (Wright et. al 1984). Increased human activities in a watershed are routinely linked to increases in fine sediments. Thus, predictions based on sediment composition may come at risk of introducing unquantifiable bias towards disturbed conditions. Our models performed comparably to other predictive models in the region, while maintaining “purity” of predictor variables.

Null model performance Performance of the Northern Basin and Range (NBR) null model cannot be assessed in the same way as predictive models. By definition, the mean O/E value for the reference sites used to build the null model is 1.0 (Table 3). Precision can be estimated by looking at the SD of O/E values for reference sites (Table 3). The high SD of O/E values for reference sites suggests low precision. Obviously, having only nine reference sites in the NBR limits our confidence in our assessment of biological condition in this region.

Using the models A separate report is available on the Xerces Society website (http://www.xerces.org/aquatic/predator/) which outlines in detail the data formatting requirements for the bug and predictor variables data files. It is crucial to follow the details of this report closely, as the software on the WCM website has very specific requirements.

PREDATOR outputs When a user submits properly formatted predictive model files to the WCM website, the predictive model software generates four output files. For a basic analysis we are primarily concerned with two files: “Site Test Results” and “O Over E”. Both the “Probability Matrix” and “Summary” files are useful for determining why a site may be disturbed (see Western Center for Monitoring 2006). (See below for further descriptions on how to use these outputs.)

Site Test Results--This file shows if a sample was within the experience of the model. This means that the predictor variables at a test site were within the range of the predictor variables at reference sites used to build the model. A chi-squared test is used to declare outliers based on the multivariate distance between a site's set of predictor values, and the values seen at reference sites. Samples outside of the experience of the reference sites used to build the model are considered outliers and are flagged as such in the “Site Test Results” file. They will still be scored, but it is up to the user to determine if the assessment is valid. It is important to note that this test is for predictor variables only.

16

Other important variables, such as year or sample abundances are not included (see the next paragraph).

An example of the potential pitfalls of assessing outlier samples that are not identified in the “Site Test Results” file is shown in Figure 3. O/E scores are shown for four groups of samples assessed by the MWCF model. “Count Outlier” samples had less than 200 individual macroinvertebrates and “Year Outlier” samples were collected in years earlier than the reference sites used to build the MWCF model. All “Test” sites were within the model bounds of bug count and sample year. Mean O/E scores from both count outlier (0.77) and year outlier (0.86) populations are lower than the mean O/E scores for the test population (0.91). Low counts (<200 bugs) affected O/E scores more than assessing a sample collected prior to 1998. This is not surprising, since fewer bugs in a sample should reduce the likelihood of collecting the expected taxa. Similarly, the bias towards lower O/E scores in year outlier samples could also be due to low numbers of macroinvertebrates. One of the major methods changes in macroinvertebrate sampling, beginning in 1999, was to increase the sub-sampling effort from 300 to 500 individuals.

To assess how PREDATOR models score data collected beyond timeframe the initial model development (post 2004), DEQ should continue to re-sample reference sites to track model scores. If reference sites used to build the model show O/E scores deviating from 1.0, then it may be necessary to re-calibrate the model with more current data.

Figure 3 illustrates the need to assess samples outside of the model experience with caution. The “Site Test Results” file will notify the user if the predictor variables for a test sample are beyond the experience of the model. However, it is the users’ responsibility to ensure their samples are in the correct regions, were sampled in the correct season, follow similar collection and processing protocols, have enough bugs, and were sampled no earlier than 1998. Failure to comply with any one of these conditions may lead to inaccurate predictions of O/E.

O Over E--This file contains model scores (O/E), as well as the number of observed reference taxa (O), and the number of expected reference taxa (E). The output includes calculations for two probability of capture (Pc) thresholds: 0 and 0.5. The PREDATOR models are based on Pc > 0.5 (this means the model uses only bugs with a greater than 50% likelihood of being collected at reference sites). Make sure you use the O/E scores associated with Pc > 0.5.

17

Figure 3. Frequency distributions of O/E scores for samples assessed by the MWCF model. Dark horizontal bands represent the mean for each group. Notches in the boxes are approximations of 95% confidence intervals (±1.58 Inter-quartile Range/(√n)). “Reference” = samples used to build the MWCF model (n=38); “Test” = samples not used to build the model, but within the experience of model constraints (n=252); “Count Outlier” = samples with less than 200 bugs (n=116), “Year Outlier” = samples collected prior to 1998 (n=133).

18

Benchmarks of biological condition For PREDATOR’s predictive models, the distribution of reference O/E scores is used to establish benchmarks for describing the biological condition of a sample (Table 4). Benchmarks for the predictive models were based on the 10th and 25th percentiles of reference distributions (Turak et. al 1999, Clarke et. al 2003, Ostermiller and Hawkins 2004). An O/E score ≤ 10th percentile of reference scores is considered in “Most disturbed” condition. An O/E score > 10th and ≤ 25th percentiles is considered in “Moderately disturbed” condition. An O/E score > 25th and ≤ 95th is considered in “Least disturbed” condition. Samples with O/E values > 95th percentile of reference sites are considered to be “Enriched”. The O/E percentile benchmarks can also be represented as % taxa loss or gain (Table 4).

Table 4. O/E benchmarks for describing biological condition for predictive PREDATOR models. (MWCF = Marine Western Coastal Forest; WC+CP = Western Cordillera + Columbia Plateau.) Biological Reference Condition MWCF WC+CP percentile Class % Common % Common O/E Taxa O/E Taxa Loss/Gain Loss/Gain Most ≤ 10th ≤ 0.85 ≥ 15% ≤ 0.78 ≥ 22% loss disturbed Moderately 0.79 – > 10th to 25th 0.86 - 0.91 9 – 14% 8 – 21% loss disturbed 0.92 Least 0 - 8% loss 0.93 – 0 - 7% loss > 25th to 95th 0.92 - 1.24 disturbed 0 - 24% gain 1.23 0 - 23% gain Enriched > 95th > 1.24 > 24 % gain > 1.23 > 23% gain

Table 5. Benchmarks for describing biological condition for the Northern Basin and Range (NBR) null PREDATOR model. % Common Biological Condition O/E Taxa Class Loss/Gain Most disturbed ≤ 0.50 ≥ 50% loss Moderately disturbed > 0.50 to ≤ 0.75 25-49% loss Least disturbed > 0.75 to 1.00 < 25% loss Enriched > 1.30 > 30% gain

19

The biological condition of enriched sites is possibly of concern due to the potential for streams to show an increase in diversity as a result of small to moderate levels of disturbance. This concept is known as the intermediate disturbance hypothesis (Ward & Stanford 1983). A high PREDATOR score may be an early warning sign that human activities are altering the macroinvertebrate assemblage, but not yet at a level that has led to assemblage degradation. Alternatively, a high PREDATOR score may simply indicate that a stream reach has exceptionally high richness, potentially representing unique communities worthy of special protection or preserve status.

The choice of these benchmarks was a trade-off between balancing errors in identifying a sample as disturbed when it truly isn’t (type I error), or failing to recognize biological disturbance when it exists (type II error) (Sokal and Rohlf 1995). Also, the choice of percentiles was made for consistency with other bioassessment work in this region. Benchmarks such as 1 or 2 standard deviations from the reference site means could also have been chosen. These standard deviations are also statistics that describe the distribution of reference scores and frequently are similar to the percentile benchmarks ultimately chosen.

Due to the low number of reference sites in the NBR (9) and the poor model performance (reference O/E SD = 0.29), the benchmarks for the NBR null model are less stringent than for the predictive models. With so few reference sites, the use of statistics representing the distribution of reference O/E scores did not make sense. Instead, percentages of taxa loss were chosen that match those used in Stoddard et. al (2005). These criteria are more conservative against type-I errors, but less protective of the resource in that type-II errors are much more likely. This points out the need for much more effort in southeastern Oregon (SEOR) to improve DEQ’s reference site network and develop better models for this region. Until DEQ develops a more accurate model for SEOR, I recommend using the SEOR null model with caution in bioassessments.

Population Assessments Example population assessments are shown for all samples assessed in the Coast Range ecoregion and the Willamette Valley ecoregion (Figure 4). The percent of samples (y- axis) which fall below or above the MWCF benchmarks (x-axis) are shown. The results represent the condition of all samples in DEQ’s database from the two ecoregions. If the sites were chosen randomly, we could make an estimate of the percent of stream miles in each biotic condition class with error estimates, rather than simply the percent of samples. This would provide an unbiased assessment with quantifiable estimates of error.

With the results shown in Figure 4, it is possible to prioritize future monitoring activities. For instance, there are a substantially lower number of samples collected in the Willamette Valley (35, compared to the Coast Range’s 217). With such a small sample size our confidence in the results are diminished. We may want to increase our assessment effort in the Willamette Valley to gain a better understanding of current conditions. (DEQ completed field sampling of a random study design in the Willamette Valley in 2005. Results of this study should be available in 2008.) In the Coast Range, we may want to go back and take additional samples at the locations of the 13% of

20

samples that are “moderately disturbed” and the 1% of samples that are “enriched”, to get a better estimate of biological condition. With repeated sampling, we can determine if these locations are significantly different from reference conditions (see “individual site assessments” below).

Correlations of PREDATOR scores to environmental variables would be useful in examining patterns of low or high biological condition. Hughes et. al (2004) found the environmental variables most strongly correlated to aquatic vertebrate IBI scores in the Coast Range ecoregion depended on the type of stream. For those streams draining more erosive sedimentary lithology, IBI scores were lower in streams with higher amounts of fine sediments. For streams draining more resistant lithology, road density was most highly associated with low IBI scores. Also shown by Hughes et. al (2004), simple maps of biological condition can be an effective way of presenting patterns of biological condition, as they found pockets of good IBI scores in regions dominated by wilderness or national parks.

Individual site assessments Assessing biological condition at a single site involves direct comparisons to the mean reference condition. In other words, is the average O/E score at a site significantly different from the reference average? Unlike population assessments where we utilize percentiles of the reference distribution to determine the percent of resource within a given quality category, here the intent is to determine if the biological condition at a single location is significantly different from reference. (It is important to note that the sampling methods employed are designed to represent the macroinvertebrate assemblage within the sampled stream reach, and not the conditions across the larger watershed or landscape. Thus, much care should be exercised in extrapolating results beyond the surveyed reach.)

If a single sample falls below the 10th percentile of the reference distribution, the sample is considered to be outside the reference distribution. We feel confident that a single sample score below the 10th percentile is not different simply by chance, but rather a true difference in biological condition exists (assuming the site is not an outlier for any reason). In this case, a single sample is sufficient to classify the stream reach as biologically disturbed, or “not supporting” the beneficial use. However, if a sample falls between the 10th and 25th percentiles of the reference distribution (“moderately disturbed”), there is less confidence that the O/E score is outside of the reference distribution. In this case, DEQ recommends repeated measures of O/E to determine if a significant difference in biological condition exists. We also recommend assessments include surveys of water quality, instream and riparian habitat, and remote sensing of the watershed (GIS) to provide insights into possible sources of disturbance. A site with a “most disturbed” O/E score and minimal signs of human influences may indicate that the site was not accurately modeled with the current set of reference sites. These are important findings that may be used to increase the future accuracy of predicting locally common reference taxa.

21

Enriched (1%) 100 A 90 Least disturbed (56%) 80

70 60 50

40 Moderately disturbed (13%) 30

Percent of Samples 20 Most disturbed (30%) 10 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4

O / E

B Enriched (2%) 100 90 Least disturbed (30%) 80 Moderately disturbed (2%) 70 60

50 40 Most disturbed (65%) 30

Percent of Samples 20 10

0

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 O / E Figure 4. The extent of biotic condition classes for samples in the Coast Range ecoregion (panel A, n=217) and the Willamette Valley ecoregion (panel B, n=35). Vertical lines are the 10th, 25th, and 95th percentiles (from left to right, respectively) of MWCF reference O/E scores (n = 217).

22

For a sample with higher than expected O/E (enriched), determining biological condition would also require further monitoring of the stream and its watershed. First, a combination of on-site (field observations) and remote sensing (GIS) screens of the watershed could be performed to identify potential sources of human disturbances. Those sites with levels of human disturbances similar to regional reference sites may be deemed naturally enriched. Those with higher levels of human activities may require further field sampling to determine if the existing activities are affecting the beneficial use. Repeated macroinvertebrate sampling, sampling for water quality, and instream and riparian habitat sampling may provide additional information concerning potential causes of disturbance.

Causes of poor biological condition Identifying biological condition does not satisfy all bioassessment needs. Knowing a site is in poor biological condition is useful, but unless we are able to identify the cause(s) of impairment, we are at a loss for how to most effectively go about improving the stream.

Information on tolerances from individual taxa can be paired with the information from PREDATOR to get a sense of the likelihood of each variable as a potential cause of poor biological condition. DEQ has developed optima and tolerances for macroinvertebrate taxa to both seasonal maximum temperature and percent fine sediments (Huff et. al 2006).

The “Probability Matrix” output from the WCM shows missing taxa (they were expected to occur, but were not collected) and replacement taxa (they were not expected to occur, but were collected). With information regarding individual taxa sensitivities (optima and tolerances) to different stressors, it is possible to tease out possible causes of disturbance. In Figure 5, two streams in most disturbed condition (PREDATOR O/E < 0.85) are shown. Poor biological condition in Lower Mill Creek may be related to temperature. The missing taxa show lower temperature optima (~16-17 °C) compared to the collected (~17-18 °C) and replacement (~16.5-21 °C). On the other hand, replacement taxa show a very broad range of fine sediment optima, both lower and higher than collected and missing taxa. At Panther Creek, poor biological condition appears to be related to fine sediments. The missing and replacement taxa optima are nearly identical, and overlap the collected taxa optima. Fine sediment optima are much higher for replacement taxa (~9-18 °C) than missing taxa (~6-9 °C) and collected taxa (~8-10 °C).

23

Figure 5. Identifying potential causes of impairment in two sites with O/E in most disturbed condition. The graphs show sites where poor biological condition is potentially related to temperature (panel A) and fine sediment (panel B). Optima for individual taxa were calculated with weighted averaging (Huff et. al 2006). The horizontal axis represents both percent fine sediments and temperature (oC).

24

The importance of assessing multiple assemblages When assessing the condition of a study area it would be wise to include the results of the conditions of multiple assemblages. Different assemblages may respond differently to various stressors. If we wanted to assess chemical water quality in a region, we would measure multiple parameters and not just dissolved oxygen. Similarly, it may be unwise to base our conclusions of the overall biological condition in a region entirely on the results of one assemblage.

An assessment of aquatic vertebrate assemblages in the Coast Range ecoregion showed 45% of stream miles in most disturbed conditions (Hughes et. al 2004), while Figure 4 shows 30% of macroinvertebrate samples are most disturbed. Direct comparisons are difficult because Hughes et. al uses a random study design and slightly different condition class benchmarks. Combining information from both assemblages would allow for more robust decisions to be made on the current status of biological condition, as well as the direction of future resource management in the Coast Range.

Not only can multiple assemblages show varying levels of the resource in disturbed condition, they can also show varying responses to stressors. DEQ (2005) assessed the biological condition of aquatic vertebrates and macroinvertebrates in the Coastal Coho Evolutionarily Significant Unit (ESU), which overlaps considerably with the Coast Range ecoregion, from 1994-2003. Both assemblages showed similar amounts of stream resource in most disturbed biological condition (28% and 36%, respectively). However, when the major stressors affecting each community were assessed (relative risk, Van Sickle et. al 2006), they found macroinvertebrates were significantly affected by high levels of fine sediment, while aquatic vertebrates showed an insignificant response.

A similar study in the Lower Columbia ESU, DEQ (2007) showed a much higher percentage of stream miles in most disturbed condition for aquatic vertebrates than macroinvertebrates (22% and 7%, respectively). Of all the stressors showing significant risks to either assemblage, only three stressors were not significantly affecting both assemblages. Low dissolved oxygen, higher human disturbances in the riparian, and lower canopy condition showed significant affects on vertebrate condition, but not on macroinvertebrate condition. The other significant stressors consistently showed ~10x greater risk to macroinvertebrate assemblages than aquatic vertebrates. These results suggest it is important to look at multiple assemblages to fully assess the amount of stream resource in peril (most disturbed condition), as well as what types of stressors are responsible for the poor biological conditions.

25

Conclusions DEQ developed three regional models (PREDATOR) that can be used to assess biotic condition of Oregon’s wadeable streams using macroinvertebrates. Two of these models are predictive models, while the third is a null model with no predictive component. We attempted to build predictive models that would be able to assess all wadeable streams in Oregon, but model performance decreased significantly when macroinvertebrate samples from reference sites in southeastern Oregon were included.

There are several important considerations for using PREDATOR models. Those wishing to utilize PREDATOR should carefully consider the implications of scoring samples taken outside of the context of the models. Failure to follow the specifications of each PREDATOR model may lead to inappropriate interpretations of biotic condition.

Those ecoregions where DEQ currently has abundant reference sites have had large scale monitoring efforts within them (Cascades, Coast Range, and Blue Mountains). However, there are several ecoregions where DEQ’s reference samples are limited (Willamette Valley, East Cascades, Columbia Plateau, Klamath Mountains, and Northern Basin and Range, Snake River Plains). Precision in assessments of biotic condition in these regions would likely be improved by increasing the monitoring effort in these regions, as well as polling neighboring states for potential reference sites.

In addition to increasing the reference site pool for those ecoregions with relatively few selected sites, DEQ should develop and maintain a long-term reference site sampling plan. We need to be able to assess whether or not biological assemblages at reference sites are stable, or if they are changing through time. If they are changing—that is O/E scores at reference sites are significantly different from those original calibration set— then we need to re-model. Without monitoring a subset of the reference sites periodically, we will not have a good understanding of how well the models work under future conditions. Given the potential for climate change to impact stream communities (e.g., decreased streamflow, increased temperature, etc.), it is important to adopt measures that allow for long-term assessment of PREDATOR’s effectiveness. Finally, it would make sense to implement a statewide network of reference sites where resource management and land activities can remain relatively consistent. If we allow increased human activities or disturbances in reference watersheds, then future measurements at these sites will likely show a degradation of reference sites’ biological condition.

Future versions of PREDATOR As new reference sites are sampled, the models will be re-examined with the intent of improving our assessments in poorly represented regions. Incorporation of more reference sites will provide additional statistical precision in O/E scores by assessing model performance using independent reference datasets. Since 2004, DEQ has collected reference macroinvertebrate samples from western Oregon, especially in the Klamath Mountains ecoregion. Additionally, DEQ should partner with neighboring states to share reference sites, not only for RIVPACS-type models, but also for establishing reference benchmarks for water chemistry and physical habitat attributes from streams and rivers.

26

Future versions of PREDATOR may change the environmental predictor variables if additional reference site data shows this is necessary. New statistical procedures, such as Random Forests and Classification and Regression Tree models (Cao et. al 2007), may allow for more precise models. Ultimately, DEQ is interested in making sure the PREDATOR models are available and easy to use for as wide an audience as possible, while achieving maximum model performance.

Recommendations and Needs

1) Periodic sampling of reference sites used to build the models.

2) Increased monitoring in southeastern Oregon and the Columbia Plateau ecoregion. (Both random surveys and targeted reference sampling.)

3) Obtain reference sites from neighboring states—especially in regions where DEQ currently lacks adequate reference data sets.

4) Resources to educate and train third parties interested in using PREDATOR.

5) Develop accurate, precise, and sensitive models or indices for algae, aquatic vertebrates, and macroinvertebrates that can assess the biological condition of all stream and rivers in Oregon.

27

Acknowledgements Dave Huff of the University of Minnesota was instrumental in building the PREDATOR models. Chuck Hawkins and Mark Vinson of Utah State University and Rob Plotnikoff of Washington Department of Ecology for providing data used to build the models. John Van Sickle of USEPA’s Office of Research and Development provided R-code which has vastly improved our ability to create and assess model performance, as well as a review of this document. Bob Hughes of Oregon State University provided an excellent critique of this paper. The Oregon Watershed Enhancement Board for funding that allowed Jeff Adams of the Xerces Society to pull together training materials.

Literature Cited

Centre for Ecology and Hydrology. 2006. www.ceh.ac.uk/sections/re/RIVPACS.html

Cao, Y., C.P. Hawkins, and MA Kosterman. 2007. Modeling natural environmental gradients improves the accuracy and precision of diatom-based indicators. Journal of the North American Benthological Society. 26(3): 566-585.

Drake, D.L. 2004. Selecting Reference Condition Sites - An Approach for Biological Criteria and Watershed Assessment. Oregon Department of Environmental Quality. http://www.deq.state.or.us/lab/Biomon/reports/WSA04-002.pdf

DEQ. 2004. Mode of Operations Manual. Oregon Department of Environmental Quality. http://www.deq.state.or.us/lab/qa/DEQ03-LAB-0036-SOP.pdf

DEQ. 2005. Oregon Coastal Coho Assessment, Section B-1: The nature and extent of threats being addressed by the conservation effort are described. Oregon Department of Environmental Quality. http://nrimp.dfw.state.or.us/OregonPlan/default.aspx?p=152&path=ftp/reports/Final%2 0Reports/Agency%20Reports/ODEQ&title=&link=

DEQ. 2007. Lower Columbia Wadeable Streams Report. Oregon Department of Environmental Quality. http://www.deq.state.or.us/lab/qa/DEQ08-LAB-003.pdf

Hawkins, C.P. 2006. Quantifying biological integrity by taxonomic completeness: evaluation of a potential indicator for use in regional and global-scale assessments. Ecological Applications. 16: 1277-1294.

Hargett, E.G., J.R. ZumBerge, C.P. Hawkins, and J.R. Olson. 2007. Development of a RIVPACS-type predictive model for bioassessment of wadeable streams in Wyoming. Ecological Indicators. 7: 807-826.

Huff D.D., S. Hubler, Y. Pan, and D. Drake. 2006. Detecting Shifts in Macroinvertebrate Community Requirements: Implicating Causes of Impairment in Streams. Oregon Department of Environmental Quality. DEQ06-LAB-0068-TR

28

Karr, J.R. and E.W. Chu. 1999. Restoring Life in Running Waters: Better Biological Monitoring. Island Press. Washington, D.C. 206pp.

McCune, B. and J.B. Grace. 2002. Analysis of Ecological Communities. MjM Software Design. Gleneden Beach, OR. 300pp.

Moss, D., J.F. Wright, M.T. Furse, and R.T. Clarke. 1999. A comparison of alternative techniques for prediction of the fauna of running water sites in Great Britain. Freshwater Biology. 41: 167-181.

Norris, R.H. and C.P. Hawkins. 2000. Monitoring river health. Hydrobiologia. 435: 5- 17.

Ostermiller, J.D. and C.P. Hawkins. 2004. Effects of sampling error on bioassessments of stream ecosystems: application to RIVPACS-type models. Journal of the North American Benthological Society. 23 (2): 363-382.

Reynoldson, T.B., R.C. Bailey, K.E. Day, and R.H. Norris. 1995. Biological guidelines for freshwater sediment based on Benthic Assessment of Sediment (the BEAST) using multivariate approaches for predicting biological state. Australian Journal of Ecology. 20: 198-219.

Reynoldson, T.B., R.H. Norris, V.H. Resh, K.E. Day, and D.M. Rosenberg. 1997. The reference condition: a comparison of multimetric and multivariate approaches to assess water-quality impairment using benthic macroinvertebrates. Journal of the North American Benthological Society.. 16: 833-852.

Rosenberg, D.M. and V.H. Resh. 1993. Freshwater Biomonitoring and Benthic Macroinvertebrates. Chapman and Hall. New York, NY. 488pp.

Simpson, J.C. and R.H. Norris. 2000. Biological assessment of river quality: development of AUSRIVAS models and outputs. Pages: 125-142 in J.F. Wright, D.W. Sutcliffe, and M.T. Furse. 2000. Assessing the biological quality of fresh waters: RIVPACS and other techniques. Freshwater Biological Association. Ambleside, U.K.

Sokal R.R. and F.J. Rohlf. 1995. Biometry: The Principles and Practice of Statistics in Biological Research. 3rd Edition. W.H. Freeman Co. New York. 8887pp.

Stoddard, J.L., D.V. Peck, S.G. Paulsen. J. Van Sickle, C.P. Hawkins, A.T. Herlihy, R.M. Hughes, P.R. Kaufmann, D.P. Larsen, G. Lomnicky, A.R. Olsen, S.A. Petersen, P.L. Ringold, and T.R. Whittier. 2005. An ecological assessment of western streams and rivers. EPA 620/R-05/005. U.S. Environmental Protection Agency. Washington, D.C.

Stoddard, J.L., D.P. Larsen, C.P. Hawkins, R.K. Johnson, and R.H. Norris. 2006. Setting expectations for the ecological condition of streams: the concept of reference condition. Ecological Applications. 16(4): 1267-1276.

29

Van Sickle J., Hawkins C.P., Larsen D.P. & Herlihy A.H. 2005. A null model for the expected macroinvertebrate assemblage in streams. Journal of the North American Benthological Society, 24, 178–191.

Van Sickle, J., D.D. Huff, and C.P. Hawkins. 2006a. Selecting discriminant function models for predicting the expected richness of aquatic macroinvertebrates. Freshwater Biology. 51: 359-372.

Van Sickle, J., J.L. Stoddard, S.G. Paulsen, A.R. Olsen. 2006b. Using relative risk to compare the effects of aquatic stressors at a regional scale. Environmental Management. 38(6): 1020-1030.

Van Sickle, J., D.P. Larsen, and C.P. Hawkins. 2007. Exclusion of rare taxa affects performance of the O/E index in bioassessments. Journal of the North American Benthological Society. 26(2): 319-331.

Wallace, J.B. and J.R. Webster. 1996. The role of macroinvertebrates in stream ecosystem function. Annu. Rev. Entomol. 41: 115-39.

Ward, J.W. and J.A. Stanford. 1983. Intermediate-Disturbance Hypothesis: An Explanation for Biotic Diversity Patterns in Lotic Ecosystems. In: Dynamics of Lotic Systems . Ann Arbor Science. Ann Arbor MI. 347-356 pp.

Western Center for Monitoring. 2006. www.cnr.usu.edu/wmc

Wright, J.F., D. Moss, P.D. Armitage, and M.T. Furse. 1984. A preliminary classification of running water sites in Great Britain based on macroinvertebrate species and the prediction of community type using environmental data. Freshwater Biology. 14: 221-256.

Wright, J.F., D.W. Sutcliffe, and M.T. Furse. 2000. Assessing the biological quality of fresh waters: RIVPACS and other techniques. Freshwater Biological Association. Ambleside, Cumbria, U.K. 373 pp.

Xerces. 2006. www.xerces.org/aquatic/predator

30 DEQ08-LAB-0048-TR

Appendix A. Table 6. OTUs and (abbreviated) phylogenetic classifications used in the November 2005 PREDATOR models. Phylum Class Order Family Subfamily Tribe OTU Annelida Hirudinea Hirudinea Annelida Oligochaeta Oligochaeta Arthropoda Arachnoidea Sarcoptiformes Oribatidae Trombidiformes Arthropoda Arachnoidea Trombidiformes "Hydracarina" Trombidiformes Arthropoda Crustacea Amphipoda Crangonyctidae Crangonyx Arthropoda Crustacea Amphipoda Crangonyctidae Stygobromus Arthropoda Crustacea Amphipoda Gammaridae Eogammarus Arthropoda Crustacea Amphipoda Gammaridae Gammarus Arthropoda Crustacea Amphipoda Gammaridae Ramellogammarus Arthropoda Crustacea Amphipoda Hyalellidae Hyalella Arthropoda Crustacea Amphipoda Talitridae Hyalella Arthropoda Crustacea Isopoda Asellidae Arthropoda Crustacea Podocopa Ostracoda Arthropoda Insecta Coleoptera Amphizoidae Amphizoa Arthropoda Insecta Coleoptera Chrysomelidae Chrysomelidae Arthropoda Insecta Coleoptera Curculionidae Curculionidae Arthropoda Insecta Coleoptera Dryopidae Helichus Arthropoda Insecta Coleoptera Dytiscidae Dytiscidae Arthropoda Insecta Coleoptera Elmidae Ampumixis Arthropoda Insecta Coleoptera Elmidae Atractelmis_wawona Arthropoda Insecta Coleoptera Elmidae Cleptelmis Arthropoda Insecta Coleoptera Elmidae Cylloepus

Phylum Class Order Family Subfamily Tribe OTU Arthropoda Insecta Coleoptera Elmidae Dubiraphia Arthropoda Insecta Coleoptera Elmidae Heterelmis Arthropoda Insecta Coleoptera Elmidae Heterlimnius Arthropoda Insecta Coleoptera Elmidae Lara Arthropoda Insecta Coleoptera Elmidae Microcylloepus Arthropoda Insecta Coleoptera Elmidae Narpus Arthropoda Insecta Coleoptera Elmidae Optioservus Arthropoda Insecta Coleoptera Elmidae Ordobrevia Arthropoda Insecta Coleoptera Elmidae Rhizelmis Arthropoda Insecta Coleoptera Elmidae Stenelmis Arthropoda Insecta Coleoptera Elmidae Zaitzevia Arthropoda Insecta Coleoptera Haliplidae Haliplidae Arthropoda Insecta Coleoptera Hydraenidae Hydraenidae Arthropoda Insecta Coleoptera Hydrochidae Hydrochus Arthropoda Insecta Coleoptera Hydrophilidae Hydrophilidae Arthropoda Insecta Coleoptera Noteridae Noteridae Arthropoda Insecta Coleoptera Psephenidae Acneus Arthropoda Insecta Coleoptera Psephenidae Dicranopselaphus Arthropoda Insecta Coleoptera Psephenidae Eubrianax_edwardsi Arthropoda Insecta Coleoptera Psephenidae Psephenus Arthropoda Insecta Coleoptera Ptilodactylidae Ptilodactylidae Arthropoda Insecta Coleoptera Scirtidae Scirtidae Arthropoda Insecta Coleoptera Staphylinidae Staphylinidae Arthropoda Insecta Diptera Athericidae Atherix

32

Phylum Class Order Family Subfamily Tribe OTU Arthropoda Insecta Diptera Blephariceridae Blephariceridae Arthropoda Insecta Diptera Ceratopogonidae Ceratopogoninae Ceratopogoninae Arthropoda Insecta Diptera Ceratopogonidae Dasyheleinae Dasyheleinae Arthropoda Insecta Diptera Ceratopogonidae Forcipomyiinae Forcipomyiinae Arthropoda Insecta Diptera Chaoboridae Chaoboridae Arthropoda Insecta Diptera Chironomidae Chironominae Chironominae Arthropoda Insecta Diptera Chironomidae Diamesinae Diamesinae Arthropoda Insecta Diptera Chironomidae Orthocladiinae Orthocladiinae Arthropoda Insecta Diptera Chironomidae Podonominae Podonominae Arthropoda Insecta Diptera Chironomidae Prodiamesinae Prodiamesinae Arthropoda Insecta Diptera Chironomidae Tanypodinae Tanypodinae Arthropoda Insecta Diptera Culicidae Culicidae Arthropoda Insecta Diptera Deuterophlebiidae Deuterophlebia Arthropoda Insecta Diptera Dixidae Dixa Arthropoda Insecta Diptera Dixidae Dixella Arthropoda Insecta Diptera Dixidae Meringodixa Arthropoda Insecta Diptera Dolichopodidae Dolichopodidae Arthropoda Insecta Diptera Empididae Chelifera_ Arthropoda Insecta Diptera Empididae Clinocera Arthropoda Insecta Diptera Empididae Hemerodromia Arthropoda Insecta Diptera Empididae Neoplasta Arthropoda Insecta Diptera Empididae Oreogeton Arthropoda Insecta Diptera Empididae Phyllodromia Arthropoda Insecta Diptera Empididae Trichoclinocera

33

Phylum Class Order Family Subfamily Tribe OTU Arthropoda Insecta Diptera Empididae Wiedemannia Arthropoda Insecta Diptera Ephydridae Ephydridae Arthropoda Insecta Diptera Muscidae Muscidae Arthropoda Insecta Diptera Pelecorhynchidae Glutops Arthropoda Insecta Diptera Phoridae Phoridae Arthropoda Insecta Diptera Psychodidae Maruina Arthropoda Insecta Diptera Psychodidae Pericoma/Telmatoscopus Arthropoda Insecta Diptera Psychodidae Psychoda Arthropoda Insecta Diptera Ptychopteridae Ptychopteridae Arthropoda Insecta Diptera Sciomyzidae Sciomyzidae Arthropoda Insecta Diptera Simuliidae Prosimulium Arthropoda Insecta Diptera Simuliidae Simulium Arthropoda Insecta Diptera Simuliidae Twinnia Arthropoda Insecta Diptera Stratiomyidae Stratiomyidae Arthropoda Insecta Diptera Syrphidae Syrphidae Arthropoda Insecta Diptera Tabanidae Tabanidae Arthropoda Insecta Diptera Tanyderidae Tanyderidae Arthropoda Insecta Diptera Thaumaleidae Thaumalea Arthropoda Insecta Diptera Tipulidae Ulomorpha Arthropoda Insecta Diptera Tipulidae Limoniinae Antocha Arthropoda Insecta Diptera Tipulidae Limoniinae Cryptolabis Arthropoda Insecta Diptera Tipulidae Limoniinae Dicranota Arthropoda Insecta Diptera Tipulidae Limoniinae Erioptera Arthropoda Insecta Diptera Tipulidae Limoniinae Gonomyia

34

Phylum Class Order Family Subfamily Tribe OTU Arthropoda Insecta Diptera Tipulidae Limoniinae Gonomyodes Arthropoda Insecta Diptera Tipulidae Limoniinae Hesperoconopa Arthropoda Insecta Diptera Tipulidae Limoniinae Hexatoma Arthropoda Insecta Diptera Tipulidae Limoniinae Limnophila Arthropoda Insecta Diptera Tipulidae Limoniinae Limonia Arthropoda Insecta Diptera Tipulidae Limoniinae Molophilus Arthropoda Insecta Diptera Tipulidae Limoniinae Ormosia Arthropoda Insecta Diptera Tipulidae Limoniinae Pedicia Arthropoda Insecta Diptera Tipulidae Limoniinae Pilaria Arthropoda Insecta Diptera Tipulidae Limoniinae Rhabdomastix Arthropoda Insecta Diptera Tipulidae Tipulinae Tipula Arthropoda Insecta Ephemeroptera Ameletidae Ameletus Arthropoda Insecta Ephemeroptera Baetidae Acentrella Arthropoda Insecta Ephemeroptera Baetidae Acerpenna Arthropoda Insecta Ephemeroptera Baetidae Apobaetis_indeprensus Arthropoda Insecta Ephemeroptera Baetidae Baetis Arthropoda Insecta Ephemeroptera Baetidae Callibaetis Arthropoda Insecta Ephemeroptera Baetidae Camelobaetidius Arthropoda Insecta Ephemeroptera Baetidae Centroptilum Arthropoda Insecta Ephemeroptera Baetidae Diphetor_hageni Arthropoda Insecta Ephemeroptera Baetidae Fallceon_quilleri Arthropoda Insecta Ephemeroptera Baetidae Heterocloeon_anoka Arthropoda Insecta Ephemeroptera Baetidae Maccaffertium_terminatum Arthropoda Insecta Ephemeroptera Baetidae Maccaffertium_vicarium

35

Phylum Class Order Family Subfamily Tribe OTU Arthropoda Insecta Ephemeroptera Baetidae Paracloeodes_minutus Arthropoda Insecta Ephemeroptera Baetidae Plauditus Arthropoda Insecta Ephemeroptera Baetidae Procloeon Arthropoda Insecta Ephemeroptera Baetidae Pseudocloeon Arthropoda Insecta Ephemeroptera Caenidae Caenis Arthropoda Insecta Ephemeroptera Ephemerellidae Eurylophella Arthropoda Insecta Ephemeroptera Ephemerellidae Ephemerellinae Attenella Arthropoda Insecta Ephemeroptera Ephemerellidae Ephemerellinae Caudatella Arthropoda Insecta Ephemeroptera Ephemerellidae Ephemerellinae Drunella_coloradensis/flavilinea Arthropoda Insecta Ephemeroptera Ephemerellidae Ephemerellinae Drunella_doddsi Arthropoda Insecta Ephemeroptera Ephemerellidae Ephemerellinae Drunella_grandis Arthropoda Insecta Ephemeroptera Ephemerellidae Ephemerellinae Drunella_pelosa Arthropoda Insecta Ephemeroptera Ephemerellidae Ephemerellinae Drunella_spinifera Arthropoda Insecta Ephemeroptera Ephemerellidae Ephemerellinae Ephemerella Arthropoda Insecta Ephemeroptera Ephemerellidae Ephemerellinae Serratella Arthropoda Insecta Ephemeroptera Ephemerellidae Timpanoginae Timpanoga_hecuba Arthropoda Insecta Ephemeroptera Ephemeridae Ephemeridae Arthropoda Insecta Ephemeroptera Heptageniidae Heptageniinae Cinygma Arthropoda Insecta Ephemeroptera Heptageniidae Heptageniinae Cinygmula Arthropoda Insecta Ephemeroptera Heptageniidae Heptageniinae Epeorus Arthropoda Insecta Ephemeroptera Heptageniidae Heptageniinae Heptagenia Arthropoda Insecta Ephemeroptera Heptageniidae Heptageniinae Ironodes Arthropoda Insecta Ephemeroptera Heptageniidae Heptageniinae Nixe/Leucocruta Arthropoda Insecta Ephemeroptera Heptageniidae Heptageniinae Rhithrogena

36

Phylum Class Order Family Subfamily Tribe OTU Arthropoda Insecta Ephemeroptera Heptageniidae Heptageniinae Stenonema Arthropoda Insecta Ephemeroptera Isonychiidae Isonychiidae Arthropoda Insecta Ephemeroptera Leptohyphidae Asioplax Arthropoda Insecta Ephemeroptera Leptophlebiidae Leptophlebiidae Arthropoda Insecta Ephemeroptera Polymitarcydae Polymitarcydae Arthropoda Insecta Ephemeroptera Siphlonuridae Siphlonuridae Arthropoda Insecta Ephemeroptera Tricorythidae Tricorythodes Arthropoda Insecta Hemiptera Belostomatidae Belostomatidae Arthropoda Insecta Lepidoptera Petrophila Arthropoda Insecta Lepidoptera Pyralidae Petrophila Arthropoda Insecta Megaloptera Corydalidae Corydalidae Arthropoda Insecta Megaloptera Sialidae Sialis Arthropoda Insecta Neuroptera Sisyridae Sisyridae Arthropoda Insecta Odonata Aeshnidae Aeshnidae Arthropoda Insecta Odonata Calopterygidae Calopterygidae Arthropoda Insecta Odonata Coenagrionidae Coenagrionidae Arthropoda Insecta Odonata Cordulegastridae Cordulegaster Arthropoda Insecta Odonata Corduliidae Corduliidae Arthropoda Insecta Odonata Gomphidae Gomphidae Arthropoda Insecta Odonata Lestidae Lestidae Arthropoda Insecta Odonata Libellulidae Libellulidae Arthropoda Insecta Plecoptera Capniidae Capniidae Arthropoda Insecta Plecoptera Capniidae Capniinae Capniidae Arthropoda Insecta Plecoptera Chloroperlidae Chloroperlinae Neaviperla

37

Phylum Class Order Family Subfamily Tribe OTU Arthropoda Insecta Plecoptera Chloroperlidae Chloroperlinae Alloperlini Alloperla Arthropoda Insecta Plecoptera Chloroperlidae Chloroperlinae Alloperlini Sweltsa Arthropoda Insecta Plecoptera Chloroperlidae Chloroperlinae Chloroperlini Haploperla Arthropoda Insecta Plecoptera Chloroperlidae Chloroperlinae Chloroperlini Plumiperla Arthropoda Insecta Plecoptera Chloroperlidae Chloroperlinae Suwallini Suwallia Arthropoda Insecta Plecoptera Chloroperlidae Paraperlinae Kathroperla Arthropoda Insecta Plecoptera Chloroperlidae Paraperlinae Paraperla Arthropoda Insecta Plecoptera Chloroperlidae Paraterlinae Utaperla Arthropoda Insecta Plecoptera Leuctridae Leuctrinae Despaxia Arthropoda Insecta Plecoptera Leuctridae Leuctrinae Leuctra Arthropoda Insecta Plecoptera Leuctridae Leuctrinae Moselia Arthropoda Insecta Plecoptera Leuctridae Leuctrinae Paraleuctra Arthropoda Insecta Plecoptera Leuctridae Leuctrinae Perlomyia Arthropoda Insecta Plecoptera Leuctridae Megaleuctrinae Megaleuctra Arthropoda Insecta Plecoptera Nemouridae Amphinemurinae Malenka Arthropoda Insecta Plecoptera Nemouridae Nemourinae Nemoura Arthropoda Insecta Plecoptera Nemouridae Nemourinae Ostrocerca Arthropoda Insecta Plecoptera Nemouridae Nemourinae Podmosta Arthropoda Insecta Plecoptera Nemouridae Nemourinae Prostoia Arthropoda Insecta Plecoptera Nemouridae Nemourinae Soyedina Arthropoda Insecta Plecoptera Nemouridae Nemourinae Visoka Arthropoda Insecta Plecoptera Nemouridae Nemourinae Zapada Arthropoda Insecta Plecoptera Peltoperlidae Peltoperlinae Sierraperla Arthropoda Insecta Plecoptera Peltoperlidae Peltoperlinae Soliperla Arthropoda Insecta Plecoptera Peltoperlidae Peltoperlinae Yoraperla

38

Phylum Class Order Family Subfamily Tribe OTU Arthropoda Insecta Plecoptera Perlidae Perlesta Arthropoda Insecta Plecoptera Perlidae Acroneuriinae Acroneurini Acroneuria Arthropoda Insecta Plecoptera Perlidae Acroneuriinae Acroneurini Calineuria Arthropoda Insecta Plecoptera Perlidae Acroneuriinae Acroneurini Doroneuria Arthropoda Insecta Plecoptera Perlidae Acroneuriinae Acroneurini Hesperoperla Arthropoda Insecta Plecoptera Perlidae Perlinae Perlini Claassenia Arthropoda Insecta Plecoptera Perlodidae Isoperlinae Calliperla Arthropoda Insecta Plecoptera Perlodidae Isoperlinae Cascadoperla Arthropoda Insecta Plecoptera Perlodidae Isoperlinae Isoperla Arthropoda Insecta Plecoptera Perlodidae Perlodinae Arcynopterygini Frisonia Arthropoda Insecta Plecoptera Perlodidae Perlodinae Arcynopterygini Megarcys Arthropoda Insecta Plecoptera Perlodidae Perlodinae Arcynopterygini Oroperla Arthropoda Insecta Plecoptera Perlodidae Perlodinae Arcynopterygini Perlinodes Arthropoda Insecta Plecoptera Perlodidae Perlodinae Arcynopterygini Setvena Arthropoda Insecta Plecoptera Perlodidae Perlodinae Arcynopterygini Skwala Arthropoda Insecta Plecoptera Perlodidae Perlodinae Diploperlini Cultus Arthropoda Insecta Plecoptera Perlodidae Perlodinae Diploperlini Kogotus Arthropoda Insecta Plecoptera Perlodidae Perlodinae Diploperlini Osobenus Arthropoda Insecta Plecoptera Perlodidae Perlodinae Diploperlini Pictetiella Arthropoda Insecta Plecoptera Perlodidae Perlodinae Diploperlini Rickera Arthropoda Insecta Plecoptera Perlodidae Perlodinae Perlodini Diura Arthropoda Insecta Plecoptera Perlodidae Perlodinae Perlodini Isogenoides Arthropoda Insecta Plecoptera Pteronarcyidae Pteronarcyinae Pteronarcellini Pteronarcella Arthropoda Insecta Plecoptera Pteronarcyidae Pteronarcyinae Pteronarcyini Pteronarcys Arthropoda Insecta Plecoptera Taeniopterygidae Taeniopterygidae Arthropoda Insecta Trichoptera Apataniidae Allomyia Arthropoda Insecta Trichoptera Apataniidae Moselyana

39

Phylum Class Order Family Subfamily Tribe OTU Arthropoda Insecta Trichoptera Apataniidae Pedomoecus Arthropoda Insecta Trichoptera Apataniidae Apataniinae Apatania Arthropoda Insecta Trichoptera Brachycentridae Amiocentrus Arthropoda Insecta Trichoptera Brachycentridae Brachycentrus Arthropoda Insecta Trichoptera Brachycentridae Micrasema Arthropoda Insecta Trichoptera Calamoceratidae Calamoceratinae Heteroplectron Arthropoda Insecta Trichoptera Glossosomatidae Agapetinae Agapetus Arthropoda Insecta Trichoptera Glossosomatidae Glossosomatinae Anagapetini Anagapetus Arthropoda Insecta Trichoptera Glossosomatidae Glossosomatinae Glossosomatini Glossosoma Arthropoda Insecta Trichoptera Glossosomatidae Protoptilinae Culoptila Arthropoda Insecta Trichoptera Glossosomatidae Protoptilinae Protoptila Arthropoda Insecta Trichoptera Goeridae Goerinae Goera Arthropoda Insecta Trichoptera Goeridae Goerinae Goeracea Arthropoda Insecta Trichoptera Helicopsychidae Helicopsyche Arthropoda Insecta Trichoptera Hydropsychidae Arctopsychinae Arctopsyche Arthropoda Insecta Trichoptera Hydropsychidae Arctopsychinae Parapsyche Arthropoda Insecta Trichoptera Hydropsychidae Hydropsychinae Cheumatopsyche Arthropoda Insecta Trichoptera Hydropsychidae Hydropsychinae Hydropsyche Arthropoda Insecta Trichoptera Hydroptilidae Ithytrichia Arthropoda Insecta Trichoptera Hydroptilidae Hydroptilinae Hydroptilini Agraylea Arthropoda Insecta Trichoptera Hydroptilidae Hydroptilinae Hydroptilini Hydroptila Arthropoda Insecta Trichoptera Hydroptilidae Hydroptilinae Hydroptilini Oxyethira Arthropoda Insecta Trichoptera Hydroptilidae Hydroptilinae Hydroptilini Paucicalcaria Arthropoda Insecta Trichoptera Hydroptilidae Hydroptilinae Leucotrichiini Leucotrichia Arthropoda Insecta Trichoptera Hydroptilidae Hydroptilinae Neotrichiini Neotrichia

40

Phylum Class Order Family Subfamily Tribe OTU Arthropoda Insecta Trichoptera Hydroptilidae Hydroptilinae Ochrottrichiini Metrichia Arthropoda Insecta Trichoptera Hydroptilidae Hydroptilinae Ochrottrichiini Ochrotrichia Arthropoda Insecta Trichoptera Hydroptilidae Hydroptilinae Stactobiini Alisotrichia Arthropoda Insecta Trichoptera Hydroptilidae Hydroptilinae Stactobiini Stactobiella Arthropoda Insecta Trichoptera Hydroptilidae Ptilocolepinae Palaegapetus Arthropoda Insecta Trichoptera Lepidostomatidae Lepidostoma Arthropoda Insecta Trichoptera Lepidostomatidae Lepidostomatinae Lepidostoma Arthropoda Insecta Trichoptera Leptoceridae Mystacides Arthropoda Insecta Trichoptera Leptoceridae Leptocerinae Athripsodini Ceraclea Arthropoda Insecta Trichoptera Leptoceridae Leptocerinae Mystacidini Mystacides Arthropoda Insecta Trichoptera Leptoceridae Leptocerinae Nectopsychini Nectopsyche Arthropoda Insecta Trichoptera Leptoceridae Leptocerinae Oecetini Oecetis Arthropoda Insecta Trichoptera Leptoceridae Leptocerinae Triaenodini Triaenodes Arthropoda Insecta Trichoptera Limnephilidae Dicosmoecinae Allocosmoecus Arthropoda Insecta Trichoptera Limnephilidae Dicosmoecinae Amphicosmoecus_canax Arthropoda Insecta Trichoptera Limnephilidae Dicosmoecinae Cryptochia Arthropoda Insecta Trichoptera Limnephilidae Dicosmoecinae Dicosmoecus Arthropoda Insecta Trichoptera Limnephilidae Dicosmoecinae Ecclisocosmoecus Arthropoda Insecta Trichoptera Limnephilidae Dicosmoecinae Ecclisomyia Arthropoda Insecta Trichoptera Limnephilidae Dicosmoecinae Onocosmoecus Arthropoda Insecta Trichoptera Limnephilidae Limnephilinae Chilostigmini Desmona Arthropoda Insecta Trichoptera Limnephilidae Limnephilinae Chilostigmini Glyphopsyche Arthropoda Insecta Trichoptera Limnephilidae Limnephilinae Chilostigmini Homophylax Arthropoda Insecta Trichoptera Limnephilidae Limnephilinae Chilostigmini Psychoglypha Arthropoda Insecta Trichoptera Limnephilidae Limnephilinae Limnephilini Asynarchus Arthropoda Insecta Trichoptera Limnephilidae Limnephilinae Limnephilini Grammotaulius

41

Phylum Class Order Family Subfamily Tribe OTU Arthropoda Insecta Trichoptera Limnephilidae Limnephilinae Limnephilini Hesperophylax Arthropoda Insecta Trichoptera Limnephilidae Limnephilinae Limnephilini Lenarchus Arthropoda Insecta Trichoptera Limnephilidae Limnephilinae Limnephilini Limnephilus Arthropoda Insecta Trichoptera Limnephilidae Limnephilinae Stenophylacini Chyrandra_centralis Arthropoda Insecta Trichoptera Limnephilidae Limnephilinae Stenophylacini Clostoeca Arthropoda Insecta Trichoptera Limnephilidae Limnephilinae Stenophylacini Hydatophylax Arthropoda Insecta Trichoptera Limnephilidae Limnephilinae Stenophylacini Philocasca Arthropoda Insecta Trichoptera Limnephilidae Pseudostenophylacinae Pseudostenophylax Arthropoda Insecta Trichoptera Odontoceridae Odontoceridae Arthropoda Insecta Trichoptera Philopotamidae Chimarrinae Chimarra Arthropoda Insecta Trichoptera Philopotamidae Philopotaminae Dolophilodes Arthropoda Insecta Trichoptera Philopotamidae Philopotaminae Wormaldia Arthropoda Insecta Trichoptera Phryganeidae Phryganeidae Arthropoda Insecta Trichoptera Polycentropodidae Polycentropodidae Arthropoda Insecta Trichoptera Psychomiidae Psychomyiinae Psychomyia Arthropoda Insecta Trichoptera Psychomiidae Psychomyiinae Tinodes Arthropoda Insecta Trichoptera Rhyacophilidae Himalopsyche Arthropoda Insecta Trichoptera Rhyacophilidae Rhyacophila Arthropoda Insecta Trichoptera Sericostomatidae Sericostomatinae Gumaga Arthropoda Insecta Trichoptera Uenoidae Thremmatinae Neophylax Arthropoda Insecta Trichoptera Uenoidae Thremmatinae Oligophlebodes Arthropoda Insecta Trichoptera Uenoidae Uenoinae Farula Arthropoda Insecta Trichoptera Uenoidae Uenoinae Neothremma Arthropoda Insecta Trichoptera Uenoidae Uenoinae Sericostriata Arthropoda Insecta Tricladida Planariidae Turbellaria

42

Phylum Class Order Family Subfamily Tribe OTU Coelenterata Hydrazoa Hydroida Hydridae Cnidaria Mollusca Bivalvia Veneroida Sphaeridae Pisidiidae Mollusca Gastropoda Basommatomorpha Planorbidae Planorbidae Mollusca Gastropoda Basommatophora Lymnaeidae Lymnaeidae Mollusca Gastropoda Basommatophora Planorbidae Planorbidae Mollusca Gastropoda Limnophila Ancylidae Ferrissia Mollusca Gastropoda Limnophila Lymnaeidae Lymnaeidae Mollusca Gastropoda Limnophila Physidae Physa Mollusca Gastropoda Limnophila Planorbidae Planorbidae Mollusca Gastropoda Mesogastropoda Hydrobiidae Hydrobiidae Mollusca Gastropoda Mesogastropoda Hydrobiidae Potamopyrgus_antipodarum Mollusca Gastropoda Mesogastropoda Pleuroceridae Juga Mollusca Gastropoda Neotaenioglossa Hydrobiidae Hydrobiidae Mollusca Pelecypoda Margaritiferidae Unionidae Mollusca Pelecypoda Sphaeriidae Pisidiidae Mollusca Pelecypoda Unionidae Unionidae Mollusca Pelecypoda Unionoida Margaritiferidae Unionidae Mollusca Pelecypoda Veneroida Sphaeriidae Pisidiidae Nematoda Nematoda Nematomorpha Nematomorpha Nemertea Prostoma Platyhelminthes Turbellaria Turbellaria

43 DEQ08-LAB-0048-TR Appendix B. Table 7. Candidate predictor variables that were examined in PREDATOR model development.

Variable Description Transformation Scale Source TEMPORAL order number of the day Julian date none Temporal date of the macroinvertebrate sample starting from 1 January SPATIAL longitudinal location of the Point, bottom Longitude none map site in decimal degrees of reach latitudinal location of the site Point, bottom Latitude none map in decimal degrees of reach elevation of the sampling Point, bottom Elevation site in meters determined by square-root BLM/NED (http://www.or.blm.gov/gis/resources) of reach querying on 30 meter DEM East of Cascade crest Categorical - http://www.epa.gov/wed/pages/ecoregions/na_eco.htm#Downloads Eastern Oregon (all Level III ecoregions east binary of the Cascades ecoregion) PHYSICAL/SIZE elevation change over the Stream Gradient mapped sampling reach square-root Reach 1:24K USGS DRG Map (Map slope) length divided by the reach length Watershed area in hectares as defined upstream Drainage Area by all water that flows Log10 USU/DEQ from bottom through the sample point of reach stream gradient multiplied by Stream Power the square root of the Log10 -- USU/DEQ drainage area CLIMATE mean annual precipitation at Annual PRISM - NRCS Dr. Christopher Daly OSU the sampling site in none Point Precipitation (http://www.ncgc.nrcs.usda.gov/branch/gdb/products/climate/index.html) millimeters Annual mean annual maximum air PRISM - NRCS Dr. Christopher Daly OSU Maximum temperature at sampling site none Point (http://www.ncgc.nrcs.usda.gov/branch/gdb/products/climate/index.html) Temperature in tenths of degrees Celsius LEVEL 2

ECOREGION Marine Western level II ecoregion - Categorical - Coastal Forest Combined level III Point http://www.epa.gov/wed/pages/ecoregions/na_eco.htm#Downloads binary Ecoregion ecoregions 1 and 3 Western level II ecoregion - Categorical - http://www.epa.gov/wed/pages/ecoregions/na_eco.htm#Downloads Cordillera Combined level III Point binary Ecoregion ecoregions 4, 78, 11, and 9 level II ecoregion - Western Interior Categorical - http://www.epa.gov/wed/pages/ecoregions/na_eco.htm#Downloads Combined level III Point Basin and Range binary ecoregions 10, 12, and 80 LEVEL 3

ECOREGION Categorical – http://www.epa.gov/wed/pages/ecoregions/na_eco.htm#Downloads Coast Range Level III ecoregion 1 Point binary Categorical – http://www.epa.gov/wed/pages/ecoregions/na_eco.htm#Downloads Willamette Valley Level III ecoregion 3 Point binary Categorical – http://www.epa.gov/wed/pages/ecoregions/na_eco.htm#Downloads Cascades Level III ecoregion 4 Point binary Eastern Cascades Categorical – http://www.epa.gov/wed/pages/ecoregions/na_eco.htm#Downloads Level III ecoregion 9 Point Slopes and binary Foothills Columbia Categorical – http://www.epa.gov/wed/pages/ecoregions/na_eco.htm#Downloads Level III ecoregion 10 Point Plateau binary Categorical – http://www.epa.gov/wed/pages/ecoregions/na_eco.htm#Downloads Blue Mountains Level III ecoregion 11 Point binary Snake River Level III ecoregion 12 Categorical – Point http://www.epa.gov/wed/pages/ecoregions/na_eco.htm#Downloads

Plains binary

Klamath Categorical – http://www.epa.gov/wed/pages/ecoregions/na_eco.htm#Downloads Level III ecoregion 78 Point Mountains binary Northern Basin Categorical - http://www.epa.gov/wed/pages/ecoregions/na_eco.htm#Downloads Level III ecoregion 80 Point and Range binary SOIL TYPE Combined lithology types: alluvium, argillite and slate, conglomerate, dune sand, felsic pyroclastic, glacial drift, interlayered meta- sedimentary, lake sediment Erodible Categorical - and playa, landslide, mafic Point Walker and MacLeod 1991 USGS Lithology binary pyroclastic, meta- sedimentary phylite and schist, mixed eugeosynchlinal, sandstone, shale and mudstone, siltstone, tuff Lithology types are Resistant combined into one of two Categorical - Point Walker and MacLeod 1991 USGS Lithology classes: binary erodible or resistant

45

Appendix C.

Table 8. MWCF reference sites and corresponding environmental data. Level IV ecoregion: “1” = Coast Range, “3” = Willamette.

Mean Mean Annual Map Level II Level IV Julian Watershed Elevation Annual Maximum Air Subbasin Site name Longitude Latitude Slope ecoregion ecoregion Date Area (ha) (m) Precipitation Temperature (%) o (mm) ( C) MWCF 1b COOS Dalton Cr. -124.3252 43.2768 258 67 12 2.2 1643 166 MWCF 1b COQUILLE Bear Cr. -124.2819 43.0206 265 283 137 2.6 1770 175 Cummins MWCF 1d ALSEA Cr. -124.098 44.2669 218 2121 24 1.1 2014 160 MWCF 1d ALSEA Big Cr. -124.1058 44.1707 207 3815 4 1.1 1944 160 Cummins MWCF 1d ALSEA Cr. -124.091 44.2664 209 2076 19 1.2 2271 164 MWCF 1d ALSEA Rock Cr. -124.0912 44.1877 208 1386 55 1.3 2189 163 Cummins MWCF 1d ALSEA Cr. -124.0672 44.2686 188 1902 56 1.1 2284 163 MWCF 1d NEHALEM Harliss Cr. -123.7186 45.6897 249 120 126 4.2 3571 148 Trib to NF MWCF 1d NEHALEM Wolf Cr. -123.3837 45.7947 236 254 358 2.3 1958 142 SILETZ- MWCF 1d YAQUINA Boulder Cr. -123.6283 44.9295 241 672 560 3.0 3762 112 WILSON- TRASK- Trib to MWCF 1d NESTUCCA Clear Cr. -123.4881 45.4748 232 59 285 19.8 2515 132 WILSON- TRASK- Trib to MWCF 1d NESTUCCA Clear Cr. -123.4877 45.4742 245 1660 269 2.2 2541 139 WILSON- TRASK- Company MWCF 1d NESTUCCA Cr. -123.7435 45.5947 216 351 162 3.9 3594 152 LOWER MWCF 1f COLUMBIA Gnat Cr. -123.4719 46.1123 252 870 565 4.8 2109 142 LOWER MWCF 1f COLUMBIA Elk Cr. -123.5371 46.0577 266 1142 327 6.2 2973 142 Trib to MWCF 1f NEHALEM Gilmore Cr. -123.5329 45.9601 176 79 233 7.0 2880 148 MWCF 1g ALSEA Flynn Cr. -123.8533 44.5393 243 203 171 3.9 2262 167 MWCF 1g ALSEA Trout Cr. -123.9527 44.4726 179 1604 36 2.6 2048 164 Upper Rock MWCF 1g COQUILLE Cr. -123.7403 43.0903 195 55 809 8.4 2185 160 Trib to WF MWCF 1g COQUILLE Brummit Cr. -123.8405 43.2112 219 37 668 8.1 2278 163 MWCF 1g COQUILLE Slater Cr. -123.7995 42.9459 267 1554 213 2.9 1666 162 MWCF 1g SILTCOOS Fiddle Cr. -123.9353 43.8856 187 822 92 1.3 2105 167 Green Leaf MWCF 1g SIUSLAW Cr. -123.636 44.1602 184 75 256 1.9 2115 171 MF NF MWCF 1g UMPQUA Smith R. -123.8208 43.8756 193 2429 112 1.4 1745 163 MWCF 1g UMPQUA Lost Cr. -123.5218 43.4595 220 1805 89 1.1 1076 179 MWCF 1g UMPQUA Halfway Cr. -123.5846 43.7485 221 202 189 3.0 1254 171 MWCF 1g UMPQUA Harvey Cr. -123.941 43.7017 186 1954 42 0.4 1998 170 MWCF 1h SIXES NF Elk R. -124.2018 42.7222 193 2301 333 1.6 2791 168 MOLALLA- MWCF 3c PUDDING Deer Cr. -122.9209 45.0363 178 437 41 2.4 1047 174 UPPER SF Berry MWCF 3c WILLAMETTE Cr. -123.2988 44.7076 187 407 124 2.8 1458 170

46

SOUTH Trib to S MWCF 3d SANTIAM Santiam R. -122.8303 44.4138 214 308 211 14.5 1325 168 MWCF 3d TUALATIN Iler Cr. -123.2686 45.5974 220 233 194 3.8 1673 150 MWCF 3d TUALATIN Roaring Cr. -123.2537 45.5673 231 859 143 3.8 1608 153 Bergholtzer MWCF 3d TUALATIN Cr. -123.2389 45.697 238 299 204 6.9 1776 154 Scoggins MWCF 3d TUALATIN Cr. -123.2645 45.5154 246 3071 149 2.7 1631 150 MWCF 3d TUALATIN Beaver Cr. -123.29 45.6676 248 2016 168 0.4 1475 149 UPPER MWCF 3d WILLAMETTE Jordan Cr. -123.3006 43.932 224 240 158 1.5 1179 176 UPPER MWCF 3d WILLAMETTE Alder Cr. -123.3491 44.5999 234 464 151 3.3 1621 166

Table 9. WC+CP reference sites and corresponding environmental data. Level IV ecoregion: “10” = Columbia Plateau, “11” = Blue Mountains, “4” = Cascades, “78” = Klamath Mountains, “9” = Eastern Cascades Slopes and Foothills. Mean Mean Map Level II Level IV Julian Area Elevation Annual Annual Subbasin Site name Longitude Latitude Slope ecoregion ecoregion Date (ha) (m) Precip. Max. Air (%) o (mm) Temp.( C) NBR (Col. California Plateau) 10a Cr. -117.34700 47.51150 225 6939 588 4.8 1751 14.8 NBR (Col. Cummings Plateau) 10f Cr. -117.67200 46.32733 279 5185 643 3.4 1674 14.1 NBR (Col. Umtanum Plateau) 10g Cr. -120.57100 46.87700 265 998 605 4.2 1382 13.8 NBR (Col. Quilomene Plateau) 10g Cr. -120.09200 47.10240 183 5011 364 2.9 937 15.6 NBR (Col. Plateau) 10g Oak Cr. -120.87520 46.72900 224 7207 694 5.9 1534 13.1 NBR (Col. Plateau) 10g Naneum Cr. -120.47310 47.13830 225 1720 788 2.5 1724 11.8 LOWER WC 11a JOHN DAY Pine Cr. -120.27586 44.90565 212 3866 754 2.9 381 15.8 UPPER JOHN Cottonwood WC 11a DAY Cr. -119.63300 44.47800 228 7508 818 2.8 330 16.9 NORTH FORK WC 11a JOHN DAY Cabin Cr. -119.36100 44.92500 247 1962 704 2.1 330 16.9 UPPER WC 11b CROOKED Allen Cr. -120.17100 44.40200 235 1542 1478 2.4 584 12.1 UPPER GRANDE Limber Jim WC 11c RONDE Cr. -118.29545 45.10703 271 94 1438 1.6 737 11.6 WALLA WC 11c WALLA Mill Cr. -118.05983 45.98912 205 9194 735 1.1 1041 13.5 UPPER JOHN EF Canyon WC 11d DAY Cr. -118.87895 44.26448 223 4123 1378 2.3 483 12.0 UPPER JOHN Reynolds WC 11d DAY Cr. -118.51200 44.42149 212 2839 1338 3.1 635 13.4 UPPER JOHN Reynolds WC 11d DAY Cr. -118.52200 44.42000 231 6035 1304 1.5 584 13.4 UPPER JOHN M. Fk. WC 11d DAY Canyon Cr. -118.77542 44.25867 215 1184 1581 10.5 686 11.3 Dutch Flat WC 11d POWDER Cr. -118.12600 44.95900 229 2338 1596 0.4 686 11.7 WC 11e WALLOWA Minam R. -117.67563 45.40631 257 44647 1035 1.1 889 10.5 Little Minam WC 11e WALLOWA R. -117.67200 45.38400 230 11240 1087 1.8 838 9.7 WC 11e POWDER Eagle Cr. -117.44000 45.04100 242 5896 1451 0.4 1092 10.0 LOWER GRANDE WC 11f RONDE Wenaha R. -117.60066 45.97762 256 49435 646 1.3 635 13.1 UPPER Little WC 11h MALHEUR Malheur R. -118.31100 44.25000 233 5358 1566 1.3 737 11.8 WC 11h UPPER Little -118.31527 44.25178 217 4246 1573 13.0 737 11.8

47

MALHEUR Malheur R. NORTH FORK WC 11l JOHN DAY Onion Cr. -118.38236 44.89782 212 608 1693 3.2 787 11.3 NORTH FORK WC 11l JOHN DAY Baldy Cr. -118.31402 44.90789 211 2531 1700 3.4 838 10.8 NORTH FORK WC 11l JOHN DAY Martin Cr. -118.54964 44.95504 197 586 1674 3.1 787 11.3 NORTH FORK NF Cable WC 11l JOHN DAY Cr. -118.66512 45.05127 198 719 1420 2.1 737 11.6 MIDDLE FORK JOHN WC 11l DAY Big Cr. -118.68610 44.77780 205 296 1867 4.2 838 10.6 NORTH FORK WC 11l JOHN DAY Big Cr. -118.60358 45.01550 204 1088 1554 1.9 787 11.7 MIDDLE FORK JOHN Upper Big WC 11l DAY Cr. -118.69614 44.78433 204 491 1831 1.8 838 11.0 SF NORTH FORK Desolation WC 11l JOHN DAY Cr. -118.67306 44.79158 238 2358 1682 6.8 838 10.6 S.F. NORTH FORK Desolation WC 11l JOHN DAY Cr. -118.68400 44.81400 248 2837 1607 1.9 787 10.6 NORTH FORK WC 11l JOHN DAY Baldy Cr. -118.30700 44.90000 228 2420 1749 1.9 838 10.8 UPPER GRANDE WC 11l RONDE Beaver Cr. -118.20870 45.14693 222 3996 1497 5.3 787 11.6 UPPER GRANDE Limber Jim WC 11l RONDE Cr. -118.53180 45.09080 199 2106 1456 2.5 737 11.8 W.F. WC 11l WALLOWA Wallowa R. -117.23500 45.22000 226 5090 1791 0.3 1600 7.8 UPPER N. Fk. GRANDE Catherine WC 11l RONDE Cr. -117.61222 45.15672 210 3826 1314 10.2 1041 9.4 WC 11l IMNAHA Imnaha R. -117.02100 45.11043 211 17208 1390 0.1 1295 10.5 WC 11l POWDER Rock Cr. -118.10343 44.88639 230 4824 1604 3.3 737 12.4 UPPER JOHN Strawberry WC 11m DAY Cr. -118.69714 44.29466 210 259 2104 10.9 1143 11.8 UPPER JOHN Strawberry WC 11m DAY Cr. -118.69052 44.30075 211 366 1970 8.2 1092 11.8 WC 11m WALLOWA N Minam R. -117.46621 45.27151 253 3162 1644 0.8 1549 8.4 E.F. Lostine WC 11m WALLOWA R. -117.34800 45.21600 229 988 2156 0.5 1702 7.5 WC 11m IMNAHA McCully Cr. -117.15300 45.21500 228 255 2370 1.7 1702 8.0 EF Eagle WC 11m POWDER Cr. -117.31893 45.15849 256 144 2118 19.1 1753 7.7 LOWER COLUMBIA- WC 4a SANDY Lady Cr. -121.83532 45.31646 241 1000 778 3.1 2311 11.1 LOWER COLUMBIA- WC 4a SANDY Lost Cr. -121.84825 45.38499 238 2011 682 4.0 2667 11.6 LOWER COLUMBIA- WC 4a SANDY Tanner Cr. -121.95207 45.62197 245 6877 57 9.0 1905 14.7 LOWER COLUMBIA- WC 4a SANDY Cast Cr. -121.85295 45.37583 215 464 708 9.4 2515 12.2

LOWER COLUMBIA- WC 4a SANDY Bull Run R. -121.88868 45.48103 250 1885 742 6.6 2870 11.6 LOWER COLUMBIA- SF Salmon WC 4a SANDY R. -121.93954 45.26962 225 3072 507 3.5 2057 13.2 LOWER COLUMBIA- WC 4a SANDY Salmon R. -121.93750 45.27337 225 21413 535 1.5 2057 13.2

48

LOWER COLUMBIA- WC 4a SANDY Fir Cr. -122.02499 45.48101 230 1431 486 5.6 2261 14.5 MIDDLE COLUMBIA- WC 4a HOOD Herman Cr. -121.83534 45.67853 196 9126 172 5.0 2362 15.6 MIDDLE COLUMBIA- Harphon WC 4a HOOD Cr. -121.76515 45.68725 197 275 104 17.4 1803 15.3 MIDDLE COLUMBIA- WC 4a HOOD Eagle Cr. -121.86972 45.59537 251 5358 251 2.9 2362 12.9 East NORTH Copeland WC 4a UMPQUA Cr. -122.52135 43.23377 233 928 732 4.3 1295 14.5 NORTH WC 4a UMPQUA Canton Cr. -122.72469 43.49055 252 1127 660 3.2 1651 15.7 NORTH WC 4a UMPQUA Canton Cr. -122.72615 43.48689 210 1690 649 1.3 1651 15.8 French WC 4a MCKENZIE Pete Cr. -122.19510 44.04191 235 8030 603 6.8 1803 14.8 WC 4a CLACKAMAS Dickey Cr. -122.05389 44.93045 234 1976 762 5.9 2108 12.7 MIDDLE Trib to FORK Goodman WC 4a WILLAMETTE Cr. -122.67695 43.82910 236 504 393 12.2 1448 16.2 MIDDLE NF MF FORK Willamette WC 4a WILLAMETTE R. -122.28875 43.88823 241 34352 638 0.7 1702 14.5 NORTH WC 4a SANTIAM Rock Cr. -122.39721 44.69876 243 2375 579 2.9 2210 14.2 WC 4a MCKENZIE Lookout Cr. -122.15895 44.23210 195 1565 725 5.3 2311 14.0 COAST FORK Trib to Bear WC 4a WILLAMETTE Cr. -122.94898 43.89032 225 419 270 5.2 1346 17.0 WC 4a MCKENZIE Marten Cr. -122.52517 44.12009 227 2585 268 2.2 1448 16.8 WC 4a MCKENZIE Cash Cr. -122.88053 44.23008 233 920 382 7.2 1499 16.2 SF McKenzie WC 4a MCKENZIE R. -122.05685 43.95979 246 13182 830 1.6 1753 12.9 Trib to SF WC 4a MCKENZIE McKenzie -122.11595 43.95167 248 263 861 23.1 1803 13.4 MOLALLA- WC 4a PUDDING Trout Cr. -122.44564 45.04075 273 2471 449 3.7 2007 14.8 SOUTH WC 4a SANTIAM Moose Cr. -122.38619 44.42754 222 3600 389 2.6 1702 15.5 WC 4a CLACKAMAS Delph Cr. -122.24392 45.26692 223 1669 328 0.6 1549 16.2 MIDDLE FORK Shortridge WC 4a WILLAMETTE Cr. -122.48579 43.73887 194 519 348 7.9 1245 16.0 SOUTH WC 4a SANTIAM Donaca Cr. -122.19130 44.51892 208 276 615 15.8 2261 13.0 SOUTH WC 4a SANTIAM Donaca Cr. -122.19020 44.51914 224 290 625 14.2 2261 13.1 Upper Hot WC 4a CLACKAMAS Springs Cr. -122.16938 44.95037 224 3999 643 1.7 2210 12.0 WC 4a MCKENZIE King Cr. -122.16800 44.16170 220 38937 435 6.2 1448 15.8 MIDDLE FORK WC 4a WILLAMETTE Bedrock Cr. -122.54414 43.97354 219 354 342 3.6 1549 16.0 WC 4a MCKENZIE Tipsoo Cr. -122.20697 44.04921 220 195 771 10.5 1803 14.8 French WC 4a MCKENZIE Pete Cr. -122.20323 44.04291 220 8110 568 4.6 1803 14.8 SOUTH WC 4a SANTIAM Keith Cr. -122.28402 44.40748 228 471 467 13.1 2007 13.8 SOUTH WC 4a SANTIAM Trout Cr. -122.34636 44.39880 228 975 389 3.9 2007 15.1 WC 4a CLACKAMAS Roaring Cr. -122.11186 45.16138 231 10603 334 5.7 1803 14.7 WC 4a CLACKAMAS Eagle Cr. -122.10135 45.29933 229 3734 494 3.0 2210 13.3 WC 4a MCKENZIE King Cr. -122.17071 44.15300 220 1186 593 6.6 1600 15.8 SOUTH WC 4a SANTIAM Egg Cr. -122.21295 44.51842 223 163 589 32.9 2311 13.0

49

COAST FORK WC 4a WILLAMETTE Cedar Cr. -122.72368 43.54734 210 600 679 9.3 1702 15.1 COAST FORK WC 4a WILLAMETTE Martin Cr. -122.71896 43.54496 210 234 754 9.9 1702 15.1 WC 4a CLACKAMAS Dickey Cr. -122.05412 44.92941 224 1969 771 5.3 2108 12.7 MIDDLE FORK WC 4a WILLAMETTE Logan Cr. -122.47800 43.95680 219 1400 435 9.7 1600 15.5 LOWER COLUMBIA- Tumbling WC 4b SANDY Cr. -121.88150 45.23283 251 1218 715 8.1 1956 11.5 LOWER COLUMBIA- WC 4b SANDY Bighorn Cr. -121.92068 45.26190 217 240 534 13.7 2007 12.5 MIDDLE COLUMBIA- Lake WC 4b HOOD Branch Cr. -121.84321 45.51650 194 1746 699 2.4 2870 11.8 MIDDLE COLUMBIA- WC 4b HOOD McGee Cr. -121.77351 45.43014 229 975 860 2.5 2718 9.2 NORTH WC 4b UMPQUA Fish Cr. -122.40954 43.09827 225 338 1513 1.1 1448 11.4 NORTH WC 4b UMPQUA Boulder Cr. -122.50880 43.33037 244 6109 703 2.5 1245 15.4 MIDDLE FORK WC 4b WILLAMETTE Black Cr. -122.17709 43.71478 243 5188 918 0.3 1651 13.8 MIDDLE FORK WC 4b WILLAMETTE Fisher Cr. -122.13559 43.86306 241 2839 823 5.2 1651 13.7 NORTH WC 4b SANTIAM Opal Cr. -122.20803 44.84536 242 2728 632 2.1 2464 12.8 SF NORTH Breitenbush WC 4b SANTIAM R. -121.93993 44.76983 243 4998 756 3.6 1854 13.4 Trib to WC 4b MCKENZIE Rebel Cr. -122.14610 44.02313 194 52 958 29.1 1854 14.3 NORTH WC 4b SANTIAM Tincup Cr. -122.28132 44.86329 213 83 736 21.4 2464 13.0 MIDDLE FORK WC 4b WILLAMETTE Eagle Cr. -122.19464 43.68558 218 1817 1135 4.3 1651 14.1 SF McKenzie WC 4b MCKENZIE R. -121.98620 43.95494 247 6480 959 2.1 1803 12.5 WC 4b CLACKAMAS Doris Cr. -122.16762 44.91611 230 67 784 21.9 2311 12.3 Hot Springs Fork Collawash WC 4b CLACKAMAS R. -122.14237 44.88839 232 648 920 3.6 2565 12.3 NORTH Little North WC 4b SANTIAM Santaim R. -122.29003 44.85315 246 8317 475 1.6 2413 13.0 WC 4b CLACKAMAS Roaring R. -121.93803 45.19645 267 1405 923 4.2 2057 11.3 Welcome WC 4b CLACKAMAS Cr. -122.04160 44.87840 272 424 849 9.7 1956 12.2 Elk Lake WC 4b CLACKAMAS Cr. -122.00845 44.88987 190 6895 719 1.9 1905 12.2 WC 4b MCKENZIE Roney Cr. -122.01967 44.10760 202 476 676 15.0 1905 13.9 MIDDLE FORK WC 4b WILLAMETTE Black Cr. -122.10088 43.70004 229 1543 1079 13.1 1702 11.4 WC 4b CLACKAMAS Battle Cr. -122.07300 44.84753 243 1536 866 3.9 2108 12.9 NORTH WC 4b SANTIAM Crag Cr. -121.88874 44.73813 265 282 933 9.9 2007 13.2 MIDDLE FORK WC 4b WILLAMETTE Fisher Cr. -122.12762 43.85665 219 2545 868 3.3 1651 13.7 WC 4b MCKENZIE Eugene Cr. -122.00891 44.05882 220 5303 840 4.9 1905 13.1 Separation WC 4b MCKENZIE Cr. -122.00392 44.12679 223 14194 698 4.2 1854 14.8 WC 4b MIDDLE NF MF -122.07191 43.87530 221 18753 894 2.4 1702 12.7

50

FORK Willamtte R. WILLAMETTE NORTH WC 4b SANTIAM Opal Cr. -122.20207 44.84289 229 2512 696 5.1 2464 12.8 NORTH Battle Axe WC 4b SANTIAM Cr. -122.16729 44.85558 229 1032 802 3.1 2667 12.4 WC 4b MCKENZIE Roney Cr. -122.01753 44.10788 220 438 702 4.2 1905 13.9 WC 4b MCKENZIE Horse Cr. -122.01840 44.07177 221 13750 792 2.5 1905 13.1 NORTH WC 4b SANTIAM Cheat Cr. -121.92055 44.70736 223 454 1023 15.5 2007 13.2 SF NORTH Breitenbush WC 4b SANTIAM R. -121.88554 44.73921 223 2148 925 3.9 2007 13.2 LOWER WC 4c DESCHUTES Mill Cr. -121.75178 44.79520 257 2054 1421 1.3 2210 11.2 LOWER COLUMBIA- Rushing WC 4c SANDY Water Cr. -121.77785 45.37188 217 241 1058 9.8 2565 7.7 MIDDLE COLUMBIA- Cold Spring WC 4c HOOD Cr. -121.60804 45.35520 250 453 1479 4.5 2210 8.6 MIDDLE COLUMBIA- Cold WC 4c HOOD Springs Cr. -121.59067 45.39689 271 1961 1107 4.7 2159 10.0 MIDDLE FORK WC 4c WILLAMETTE Bear Cr. -122.20504 43.55441 234 914 1382 15.9 1651 13.1 NORTH Trib to WC 4c SANTIAM Marion Cr. -121.93481 44.59246 211 503 806 1.7 1854 13.5 MIDDLE FORK WC 4c WILLAMETTE Shady Cr. -122.18884 43.61984 238 73 1448 30.9 1600 13.7 MIDDLE FORK Gold Lake WC 4c WILLAMETTE Cr. -122.03097 43.64372 245 3664 1466 0.3 1753 11.2 WC 4c CLACKAMAS Slow Cr. -121.77675 44.90098 224 376 1304 1.2 1905 10.4 NORTH NF Santiam WC 4c SANTIAM R. -121.92270 44.49514 223 2024 1360 5.0 2057 11.6 MIDDLE COLUMBIA- Trib to WC 4d HOOD Polallie Cr. -121.63345 45.38656 226 228 1702 18.1 3175 10.0 UPPER KLAMATH WC 4e LAKE Rock Cr. -122.12968 42.56007 185 2438 1487 1.8 1346 12.5 NORTH WC 4e UMPQUA Lake Cr. -122.17159 43.25243 224 15483 1376 1.6 1194 13.2 NORTH Clearwater WC 4e UMPQUA R. -122.22998 43.24455 201 1303 1264 2.6 1194 13.3 UPPER SF Rogue WC 4f ROGUE R. -122.27775 42.55974 209 4120 1432 2.6 1346 11.9 SOUTH Castle Rock WC 4f UMPQUA Cr. -122.51800 43.11322 210 4437 835 8.1 1295 14.0 SOUTH Fish Lake WC 4f UMPQUA Cr. -122.55415 43.08846 210 2809 753 1.9 1245 14.0 SOUTH WC 4f UMPQUA Squaw Cr. -122.65500 42.94500 209 2761 732 5.0 1194 13.0 SOUTH Donegan WC 4f UMPQUA Cr. -122.65566 42.94481 209 1763 710 3.6 1194 13.0 UPPER Upper WC 4f ROGUE Bitterlick Cr. -122.64662 42.82720 203 3140 747 2.9 1194 13.4 UPPER WC 4f ROGUE Big Ben Cr. -122.31623 42.63752 209 2645 1253 10.5 1194 12.8 WC 78d CHETCO Chetco R. -123.91227 42.17385 256 2627 625 3.0 3429 16.6 WC 78e APPLEGATE Osier Cr. -123.24006 42.07481 200 83 1111 33.8 1295 13.1 LOWER WC 78e ROGUE Rueben Cr. -123.57204 42.65526 206 2527 236 4.6 1041 15.6 LOWER WC 78e ROGUE Whisky Cr. -123.63248 42.66427 269 3786 242 6.3 1041 15.8 MIDDLE WF WC 78e ROGUE Ashland Cr. -122.71851 42.14356 201 2556 969 6.5 838 13.3

51

MIDDLE WF WC 78e ROGUE Ashland Cr. -122.71527 42.14609 208 129 970 5.1 838 13.3 SMITH CAL NF Smith WC 78f OR R. -123.98200 42.04240 184 9158 433 1.8 2565 16.4 SMITH CAL Left Fork WC 78f OR Chrome Cr. -123.97130 42.05301 185 3876 416 4.2 2718 16.7 CHETCO.CAL EF IFORNIA Winchuck WC 78f OREGON R. -124.09119 42.05009 236 3344 78 0.9 2261 16.3 LOWER Shasta WC 78f ROGUE Costa Cr. -124.03501 42.57303 238 8800 52 0.6 1905 20.1 LOWER Little WC 9b DESCHUTES Badger Cr. -121.44904 45.30092 192 190 1222 8.7 940 12.0 LOWER WC 9b DESCHUTES Tygh Cr. -121.37563 45.31173 226 1368 794 10.4 584 13.9 MIDDLE COLUMBIA- WC 9b HOOD SF Mill Cr. -121.45411 45.47498 195 1468 766 3.4 1245 12.8 MIDDLE COLUMBIA- WC 9b HOOD SF Mill Cr. -121.44268 45.47860 196 3611 734 3.6 1143 12.8 LOWER WC 9d DESCHUTES Shitike Cr. -121.65073 44.74714 256 5373 1105 0.7 1143 12.3 UPPER WC 9d DESCHUTES Candle Cr. -121.68267 44.58685 224 1168 957 0.5 737 13.0 WC 9e WILLIAMSON Miller Cr. -121.93698 43.21295 184 3284 1693 0.9 991 12.0 SUMMER WF Silver WC 9e LAKE Cr. -121.25880 42.94294 228 630 1914 3.5 1041 11.5 SUMMER WC 9e LAKE Buck Cr. -121.29843 43.00178 186 2340 1784 2.0 991 11.7 LITTLE WC 9f DESCHUTES Paulina Cr. -121.42500 43.72500 175 5992 1309 0.7 483 13.3 WC 9h LAKE ABERT Dairy Cr. -120.75070 42.48855 188 3985 1779 1.2 737 12.3

52